Anda di halaman 1dari 69

Breakout Trading Technique article collections: BASIC and ADVANCED

" Technical tool insight: Price breakout"


BY ACTIVE TRADER STAFF (Active Trader, March 2001) .............................. 2
"More bang for your buck: Patterns within patterns"
BY ACTIVE TRADER STAFF (Active Trader, October 2000) ............................. 5
"Anticipating breakouts and beating slippage"
BY STEVE WENDLANDT (Active Trader, August 2000) ............................... 9
"Trading System Lab: 100-20 channel breakout system"
BY DION KURCZEK (Active Trader, June 2003) . . . . . . . .......................... 13
"Futures Trading System Lab: 100-20 channel breakout system"
BY DION KURCZEK (Active Trader, June 2003) . . . . . . . . . . . ...................... 16
"Futures Trading System Lab: 60-minute breakout system"
BY VOLKER KNAPP (Active Trader, January 2004) . . . . . . . ......................... 18
"Futures Trading System Lab: Four-percent breakout system"
BY VOLKER KNAPP (Active Trader, September 2004) . . . . . . . ....................... 20
"Broadening patterns: Clues to breakout direction"
BY THOMAS N. BULKOWSKI (Active Trader, April 2004) ............................. 23
"High, tight flag helps squeeze out profits"
BY THOMAS N. BULKOWSKI (Active Trader, December 2004) .......................... 28
"Mastering two-minute breakouts"
BY KEN CALHOUN (Active Trader, September 2001) .............................. 32
"Swing trading 10-day channel breakouts"
BY KEN CALHOUN (Active Trader, March 2002) ................................. 35
"Trading System Lab: Volatility breakout system"
BY THOMAS STRIDSMAN (Active Trader, October 2002) ............................. 39
"Futures Trading System Lab: Futures volatility breakout system"
BY THOMAS STRIDSMAN (Active Trader, October 2002) . . . . . . . . ..................... 41
"Better breakout trading: The noise channel system"
BY DENNIS MEYERS, PH.D. (Active Trader, September 2001) .......................... 42
"The long and short of it: The noise channel breakout system 2"
BY DENNIS MEYERS, PH.D (Active Trader, October 2001) . . . . . . ...................... 47
"The multibar range breakout system"
BY DENNIS MEYERS, PH.D (Active Trader, January 2004) ............................ 53
"Trading System Lab: DeMark variation"
BY THOMAS STRIDSMAN (Active Trader, September 2001) ........................... 58
"Trading System Lab: Dynamic breakout system"
BY THOMAS STRIDSMAN (Active Trader, February 2003) ............................ 60
"Futures Trading System Lab: Dynamic breakout system"
BY THOMAS STRIDSMAN (Active Trader, February 2003) . . .......................... 62
"Futures Trading System Lab: Experimenting with exits"
By VOLKER KNAPP (Active Trader, June 2004) . . . . . . ........................... 64
"Futures Trading System Lab: Monthly breakout"
BY DION KURCZEK AND VOLKER KNAPP (Active Trader, March 2004) . . . . . . . . . . . . . . . . . . . . . . . . . . 66
"Trading System Lab: 60-minute breakout system"
BY VOLKER KNAPP (Active Trader, January 2004) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
ACTIVE TRADER • www.activetradermag.com 1
TRADING Basics

Technical tool insight: Price breakout


Price breakouts are the basis of many of the most successful
trading approaches. We explain the basics of this trading technique.

T he price “breakout” is one of the


simplest — and most powerful —
concepts in trading. It occurs
when price moves forcefully out of a
consolidation or trading range (a period
The four-week highs or lows simply
represent natural resistance and support
levels.
This kind of trading system is often
referred to as stop-and-reverse (SAR),
ished its effectiveness to the point that
many traders look for false breakouts
(when price pushes through a breakout
level, only to reverse back through it) at
these levels, to take positions against the
of relatively narrow, sideways price because when a trade signal is generated, direction of the initial breakout (referred
movement) or pushes above or below an the existing position is liquidated to as “fading” the breakout).
established price level (support or resist- (stopped out) and a new position (a Breakouts are not limited strictly to
ance), initiating either temporary follow- reverse of the previous one) is established. moves to new highs of a certain number
through or a sustained trend. This basic trading rule — which gained of bars (i.e., 10-bar, 20-bar or 40-bar
The act of pushing to new highs or widespread popularity as the “20-day breakouts). As mentioned, price can also
lows (especially if the price level in ques- breakout” — was integral to many popu- “break out” through the support and
tion has been repeatedly tested in the lar mechanized trading strategies, most resistance levels of trading ranges, or
past) is evidence of strong momentum famously those of futures trader Richard other past technical milestones such as
and suggests the market has the poten- Dennis group of trend-followers known long-standing highs or lows.
tial to continue in that direction. In other as the “Turtles.” Trend-following traders Figure 1 shows 40-day breakout levels
words, the basic logic behind price (especially in the futures markets) used on a daily chart. Figure 2 shows 20-bar
breakouts is that a market making new this simple technique, or a variation of it, breakout levels on a 10-minute chart.
highs (and with potential for further to exploit strong trends in the 1970s and Figure 3 shows a breakout above the
price gain) is exhibiting strength and ’80s. However, the widespread populari- resistance level defined by a past signifi-
should be bought, while a market mak- ty of the 20-day breakout level has dimin- cant high.
ing new lows (and with potential for fur-
ther price decline) is exhibiting weak- FIGURE 1 DONCHIAN BREAKOUT CHANNELS, DAILY
ness and should be sold.
For example, the reason new 52-week 40-day Donchian breakout levels, both high and low. A basic breakout
highs or lows in stocks are so commonly approach is to buy when price exceeds the n-bar (in this case, the 40-day)
referenced is because of the implied sig- high and sell when it falls below the n-bar low.
nificance of price breaking through these Oracle Corporation (ORCL), daily
levels. This concept of price movement is Highest price
of last 40 bars 45
valid on intraday time frames as well as
daily or monthly ones.
40
Donchian breakout levels 37.00
The term “breakout” is often associated 35
with Richard Donchian, the first person
to popularize the systematic use of 30
breakout levels. His basic approach was
called the Donchian “four-week rule,” 26 9⁄16
which consisted of the following: 25
1. Go long (and cover short positions) Lowest price 21.50
when the market makes a new four- of last 40 bars 20
week high (that is, when price exceeds
the highest price of the previous four
15
weeks).
2. Go short (and cover long positions)
when the market makes a new four- 27 3 10 24 31 7 14 28 6 13 20 27 3 10 24 1 8 15 22 30 5 12 19 26 3 10 17 24 31 7 14 21 28 5 11 18 25 2 9 16 23 30 6 13 27 4
week low (that is, when price drops Jan. 2000 Feb. Mar. Apr. May June July Aug. Sept. Oct. Nov. Dec.
below the lowest price of the previous Source: QCharts by Quote.com
four weeks).

2 www.activetradermag.com • March 2001 • ACTIVE TRADER


Glossary
FIGURE 2 DONCHIAN BREAKOUT CHANNELS, INTRADAY
A false breakout occurs when price
The breakout concept is applicable to any time frame. Here, the highest pushes through a support or resist-
20-bar highs and lowest 20-bar lows are shown by the channel lines. ance level in the anticipated direc-
Oracle Corporation (ORCL), 10-minute
tion, suggesting a new price thrust or
27.25
27
trend, only to (relatively) quickly
reverse direction when no real follow-
26 9⁄16
through materializes. Because traders
26 who bought or sold on the initial
breakout may all scramble at once to
get out of their trades when the mar-
Highest price 25 ket fails to follow through, the rever-
of last 20 bars sal can be quite forceful. For this rea-
son, contrarian traders sometimes
24 fade initial breakouts to capitalize on
23.81 these short-term reversals.

23 Stop-and-reverse (SAR) refers to a


trading approach that is always in the
market, long or short. The existing
Lowest price
22 position is liquidated (stopped out) and
of last 20 bars a new position (a reverse of the previ-
ous one) is established, using the same
14 15 10 11 12 13 14 15 10 11 12 13 14 15 10 11 12 13 14 15 10
11/28 Tuesday 11/29 Wednesday 11/30 Thursday 12/1 Friday signal in the opposite direction. For
Source: QCharts by Quote.com example, a simple 40-day SAR break-
out system would buy when price
The Donchian-type breakout is also ing range, traders who go long on the exceeds the highest high of the last 40
commonly referred to as a “price chan- breakout can place protective stops in a days and sell when price falls below
nel” breakout. number of technically logical places, in the lowest low of the last 40 days.
relation to the range. First, the stop
Application could be placed below the low of the Support and resistance. Support is a
Traders using breakouts are basing their trading range. Second, a more conserva- price level that acts as a “floor,”
trades on the following principle: If tive stop placement would be in the preventing prices from dropping
price momentum is strong enough middle of the trading range (or in the below that level. Resistance is the
(either up or down) to push through a upper 25 percent of the trading range, opposite: a price level that acts as a
significant technical level, there is a etc.). Finally, the most conservative alter- “ceiling;” a barrier that prevents
good chance price will continue in that native is a stop just below the original prices from rising higher.
direction for at least a while. As a result, breakout level, which might be used by
these price levels represent logical trade
entry and exit levels with well-defined FIGURE 3 BREAKOUT ABOVE PRIOR HIGH
risk, both for traders who expect follow A prior high creates a resistance level that is tested multiple times before
through in the direction of the breakout price breaks out to the upside. A significant trending move follows.
and, as will be described shortly, traders
who are looking to fade breakouts. Sun Microsystems Inc. (SUNW), Weekly
28
Key points
25 59⁄64
Price breakouts are typically used as
trend-following signals. The greater the 24
number of days (or price bars) used to
determine the breakout, the longer-term
Breakout above 20
trend the trading system will reflect and
attempt to exploit. For example, a 20-day previously tested high
(or 20-bar) breakout would capture short-
16
er trends than a 40-day breakout, which
in turn would reflect shorter trends than
an 80-day breakout. Generally, in terms of
12
trend-following approaches, the longer-
term the breakout, the more significant
the price move and the greater the likeli- 8
hood of sustained follow through.
Breakout trading can also simplify
risk control because stop-loss levels are Jan. 1997 Apr. July Oct. Jan. 1998 Apr. July Oct. Jan. 1999
often easy to identify. For example, if
price breaks out of the upside of a trad- Source: QCharts by Quote.com

ACTIVE TRADER • March 2001 • www.activetradermag.com 3


FIGURE 4 TRADING RANGE BREAKOUT WITH STOP LEVELS
The boundaries of a trading range provide logical stop levels for a breakout trading range and possible stop points.
trade. After a downside breakout of the range, a trader, depending on how Figure 5 shows the reverse situation.
conservative he was, could place a stop-loss order at the original breakout The stock first breaks out to the down-
level, the midpoint of the range (or some other point within the range) or side of the trading range, but this turns
the upper level of the range. out to be a false breakout. The stock
reverses back into the trading range and
American Express Inc. (AXP), 2-minute eventually breaks out through the
55 upside of the trading range. Again, the
Far side of boundaries (and the midpoint) of the
trading range trading range provide logical stop levels
(stop 1)
— both for the initial downside breakout
Midpoint 54
and the subsequent upside breakout.
(stop 2)
Because of the possibility of false
moves at popular breakout levels,
53 traders looking to capture trending
moves sometimes use confirming sig-
nals to improve the likelihood of success.
Breakout level For example, after an initial upside
Trading range (stop 3) 51 15⁄16 breakout, the trader may wait for the
market to stay above the breakout level
(or close above it) for a certain number of
51 bars, or penetrate it by a certain percent-
age. Such techniques delay entry and
limit profit potential (and will result in
some missed trades), but they can also
cut down on false signals.
10:00 10:30 11:00 11:30 12:00 12:30 13:00 13:30 14:00 14:30 15:00 15:30 9:30 10:00 10:30 11:00
11/21 Tuesday
Source: QCharts by Quote.com Bottom line
The breakout concept is one of the most
a very short-term trader. reason for a trade is nullified, that posi- important in technical trading. Buying
All these choices have one thing in tion should be eliminated. (Note also, markets showing strength (upside
common: The placement of the stop cor- the second and third options would be breakouts) with further potential for
responds to a price move that negates likely short entry points for traders look- upside movement, and selling markets
the validity (to varying degrees) of the ing to fade the upside breakout.) Figure showing weakness (downside break-
original breakout. Whenever the original 4 shows a downside breakout out of a outs) with further potential for down-
side movement is the basis of many trad-
FIGURE 5 FALSE BREAKOUT AND REVERSAL ing plans and systems on many time
frames. Similarly, false breakouts are the
In this case, the stock first breaks out below the bottom of the trading
foundation of some counter-trend trad-
range, only to reverse back into the trading range and eventually break out
ing techniques. The breakout concept is
through the top of the range. In either case, the stop-loss levels are again
also easily mechanized for traders inter-
easily identified.
ested in a systematic approach. 
Microsoft Corporation (MSFT), daily 38
Additional research:
36
35 23⁄64 Trading for a Living
by Alexander Elder
34
John Wiley & Sons, 1993
32 Trading Systems and Methods
Far side of by Perry Kaufman
trading range 30 3rd edition, John Wiley & Sons, 1998
Trading range Technical Analysis of the Financial
28
Markets
26 by John Murphy
Midpoint New York Institute of Finance, 1999
24
Street Smarts
by Linda Raschke
22
False Original and Laurence A. Connors
breakout breakout level 20 M. Gordon Publishing Group, 1995
Schwager on Futures:
25 2 9 16 23 30 6 13 20 27 3 10 18 24 3 10 17 24 31 7 14 21 28 5 12 19 27 2 9 16 23 30 7 14 21
Technical Analysis
Dec. Jan. 1997 Feb. Mar. Apr. May June July by Jack Schwager
John Wiley & Sons, 1996
Source: QCharts by Quote.com

4 www.activetradermag.com • March 2001 • ACTIVE TRADER


TRADING Strategies

More bang for your buck:


PATTERNS
WITHIN PATTERNS

W hat makes a good


trade? Well, in retro-
spect, most traders
would say a nice prof-
it makes a good trade. But when you’re
putting a position on, the outcome is
unpredictable. We’d all like to know a
that determines when and where you’ll
exit with either a loss or a profit, you try
to structure a trade where the potential
reward is greater than the known risk.
The advantage of trading breakouts of
congestion patterns such as trading
ranges, triangles, flags and pennants is
trade will be good in advance, but alas, that these formations allow you to clear-
the markets are not so accommodating. ly define the risk on your trades. For
What you look for when you’re get- example, if a stock moves into a trading
ting in a trade is an entry point where range after a rally, you may look to buy
the odds of a move in your favor are bet- an upside breakout of the range in antic-
ter than average. Then, by having a plan ipation of a continuation of the uptrend.
The logical place to put an initial protec-
tive stop is below the low of the trading
FIGURE 1 FALSE BREAKOUT range, because a downside reversal
A trading range develops in the aftermath of a sharp rally. After an initial through the support of the range would
upside breakout, the stock reverses to the downside, stopping out the long be a bearish development.
position. Figure 1 provides an example. In late
June, Microsoft (MSFT) established a rel-
Microsoft Corporation (MSFT), daily atively narrow trading range after
Upside 82 approximately a 16-point rally. The stock
breakout broke out of the upside of the range
80 (around 80 1⁄8) on July 6. The initial pro-
tective stop would have been placed just
78 below the support level of the trading
range, around 76 1⁄2. A move back below
76 this level would suggest the upside
thrust was actually a false breakout and
Stopped out that the trade should be exited.
Support level used 74
as initial stop That’s exactly what happened. Two
72 5⁄8 days after entry the stock had pulled
72
back into the trading range. It moved
sideways to lower over the next several
70
days before, on July 19, penetrating the
downside of the range and stopping out
68
the long trade.
The risk on this trade was a moderate
3 5⁄8 points. But what do you do when a
12 19 26 3 10 17 24 31 7
July Aug. trading range is much wider and a stop
based on either the support or resistance
Source: Qcharts by Quote.com
level represents too large a risk? Figure 2

5 www.activetradermag.com • October 2000 • ACTIVE TRADER


How to create trade
opportunities with
increased reward FIGURE 2 RANGE RISK
and decreased risk by Using the opposite side of a trading range as a stop for a breakout trade can
result in large initial risk if the trading range is wide.
trading patterns within 130
International Business Machine Corp. (IBM), daily
patterns. 125

120 15⁄16
120

115

shows a much more volatile trading range


110
than that in Figure 1. Using the same
approach as in the previous example —
105
buying on an upside breakout of the trad-
ing range and placing an initial protective
100
stop below the low of the range — would
represent considerable risk.
95
As a result, some traders place the ini-
tial stop in the middle of the trading
90
range. This more conservative method is
4 11 18 25 1 8 15 29 6 13 27 3 10 24 31 7 14 28 6 13 20 27 3 10 24 1 8 15 22 30 5 12 19 26 3 10 17 24 31 7 14
based on the idea that a strong breakout
Nov. Dec. Jan. 2000 Feb. Mar. Apr. May June July Aug.
move should follow through immediate-
ly and not reverse back into the trading Source: Qcharts by Quote.com

range. Another way to reduce risk on


FIGURE 3 CONGESTION WITHIN CONGESTION
breakout trades is to look for shorter-
A shorter, narrower trading range forms just at the resistance level of a larg- term patterns within larger patterns that
er range. Using the support level of the smaller range as a protective level allow you to place your initial stop-loss
for an upside breakout substantially reduces the trade’s initial risk. closer to your entry point.

Oracle Corporation (ORCL), daily 90 Patterns within patterns


When the risk implied by a particular
trading range is exceptionally large, you
80
77 can look for smaller congestion patterns
near the support or resistance levels of
Wider trading range
70 the range. Basing entry and stop points
on the levels defined by the smaller pat-
tern can reduce the risk on the trade as

{
{ 60
well as provide the opportunity for early
entry into the position.
Narrow range
50 Figure 3 shows the formation of a
wide trading range in Oracle (ORCL) at
the beginning of this year. A trader look-
40
ing to enter long on an upside breakout
of this range would have to accept a risk
30 of more than 16 points, assuming the
bottom of the range was used for the ini-
tial stop-loss.
4 11 18 25 1 8 15 22 29 6 13 20 27 3 10 18 24 31 7 14 22 29 6 13 20 27 3 10 17 24 1 8 15 However, a much narrower trading
Oct. Nov. Dec. Jan. 2000 Feb. Mar. Apr. May range developed in February. Using this
Source: Qcharts by Quote.com range as the basis of an upside breakout
trade would have offered the same entry

ACTIVE TRADER • October 2000 • www.activetradermag.com 6


The advantage
FIGURE 4 FLAG NEAR RESISTANCE of trading breakouts
A small flag forms just below a well-defined resistance level, offering early
entry into the upside thrust move. of congestion
EMC Corporation (EMC), daily patterns such as
90 1⁄2

Resistance
trading ranges,
80
triangles, flags
Flag
70 and pennants is that
60
these formations
allow you to clearly
50
define the risk on
40
your trades.
30 point but a much closer stop. In this
27 4 11 18 25 1 8 15 29 6 13 27 3 10 24 31 7 14 28 6 13 20 27 3 10 24 1 8 15 22 30 5 12 19 26 3 10 17 24 31 7 case, placing a stop one tick below the
Oct. Nov. Dec. Jan. 2000 Feb. Mar. Apr. May June July Aug. low of the narrower trading range
Source: Qcharts by Quote.com would have reduced the risk to 6 3⁄4
points. For a short-term trader, this rep-
resents a large stop, but it’s still a dra-
FIGURE 5 NARROW FLAG matic improvement and the profit
A narrow flag consolidation forms near the resistance level of an intraday potential for the move out of the larger
head-and-shoulders bottom pattern. The low of the flag provides a lower-risk trading range is still intact. (Later, we’ll
stop level than the most recent swing low. look at the practical risk-reward impact
this can have on a trade.)
Nasdaq 100 Index (QQQ), 15-minute Figure 4 provides another example. In
93 3⁄8
this case, EMC Corp. (EMC) repeatedly
92 pulled back from resistance around 72 1⁄2.
Because a well-defined horizontal trad-
Resistance ing range did not develop (the stock
88 swung back and forth in an increasingly
wider range), the most recent swing low
around 51 would be the reference point
84 for the initial stop-loss — a risk of more
than 20 points.
Narrow However, as the stock bounced off that
flag low and made another run at the resist-
80
ance level, it formed a flag consolidation
from June 7 to June 12 with a high around
69 7⁄8 (the highs of the bars in the flags
76 were within 1⁄16 of each other) and a low
S S
around 66 13⁄16. The upside breakout of this
flag provided an early entry to the subse-
H quent surge that pushed the stock past
19 22 23 24 25 26 30 31 1 2 the 72 1⁄2 resistance level to new highs.
May June
Figure 5 shows a 15-minute chart of
Source: Qcharts by Quote.com the Nasdaq 100 tracking stock (QQQ).
The stock formed a large bottoming pat-

7 www.activetradermag.com • October 2000 • ACTIVE TRADER


FIGURE 6 EARLY ENTRY
tern (a head-and-shoulders bottom pat- A flag forms in the middle of a larger trading range. Even though price
tern; the preceding sell-off is not shown) gapped above the flag, playing the upside of this smaller pattern offered
with resistance around 82 5⁄8. As the stock early entry and a tighter stop on a long-side trade.
approached the resistance level for the
second time, on May 30, it consolidated in Motorola, Inc. (MOT), daily 52
a narrow flag pattern with resistance 49 171⁄256
around 82 7⁄32 and support around 81 5⁄8.
48
Playing an upside breakout of this pat-
tern and using its support level for the
initial stop (rather than the most recent 44
swing low around 76) reduced the risk on
a long trade to less than a point. Trading range
A final example is shown in Figure 6. 40
Here, in the middle of a larger trading
range with resistance around 32 3⁄8, 36
Motorola (MOT) formed a flag consolida-
tion in late-October 1999 that offered the
opportunity to trade an upside move with 32
lower risk. The stock gapped out of the Flag
flag (a bullish sign) above 31 1⁄2 and contin-
28
ued to run past the resistance of the larger
trading range. Placing a stop just below
the flag support at 29 15⁄16 would have 3 10 17 24 1 7 14 21 28 6 12 19 26 2 9 16 23 30 7 13 20 27 4 11 18 26 1 8 15 22 29 6 13 20 27 3
reduced the initial risk on the trade to less May June July Aug. Sept. Oct. Nov. Dec. Jan. 2000
than two points. As was the case with Source: Qcharts by Quote.com
Figure 4, the smaller pattern allowed you
to both use a tighter stop and get in earli- Using the measured move approach lishment of a trade with a price target
er on an upside breakout. on the smaller price swing from Jan. 28 based on the larger, longer-term price pat-
low of 46 5⁄8 to the Feb. 14 high of 64 3⁄4 (18 1⁄8 tern with a risk based on the smaller,
Structuring a trade points) sets up a shorter-term price target shorter-term price pattern.
Figure 3 provides a good example of how of 77 7⁄16. This level would mark a good Another general advantage of this
this approach can work in the context of a spot to take at least partial profits on the approach is that it increases your flexi-
complete trade plan. The rally from the position and raise the stop on the balance bility. Even if you are stopped out on a
late-October 1999 low to the early- of the position. The stock actually formed move through the support of the smaller
congestion pattern, you can still re-enter
a long position if the market reverses
When the risk implied by a particular again and breaks out above resistance a
second time. For example, a trader who
went long on the intraday upside thrust
trading range is exceptionally large, above resistance (say, at 62 5⁄8) on Feb. 14
and used the low of the smaller trading
you can look for smaller congestion patterns range (around 58 5⁄8) as the stop level,
would have been stopped out on the
near the support or resistance levels of intraday downside thrust on Feb. 22.
However, as mentioned earlier, this loss
is much smaller than the one that would
the range. have occurred had the stop been placed
below the low of the larger trading
January 2000 high was 41 7⁄32. The stock another flag after hitting a high of 76 1⁄2 on range, which was nearly 12 points lower.
then moved sideways, forming the larger Feb. 28. This consolidation marked an These patterns may develop relatively
trading range. A trader looking to buy on opportunity to exit part of the position infrequently, but they fulfill the primary
an upside breakout of the range could use with a profit; the stop on the remainder of goals of smart trading: They allow you
the measured move approach, whereby the the position could then be moved up to to establish trades with shorter-term risk
size of the previous price move is added the breakeven point, locking in a profit on and longer-term profit potential. In
to the current price, to project a price tar- the trade. (For more information on tak- future articles we’ll expand on these
get. Adding the size of the price move ing profits and moving stops, see ideas by looking at additional measuring
preceding the trading range to the low of “Opening day opportunities,” p. 42.) The objectives and ways to put breakouts
the larger trading range (around 46 5⁄8) bottom line: The development of the into context in relation to underlying
results in an upside target of 87 27⁄32. smaller trading range allowed the estab- trends of different magnitudes. 

ACTIVE TRADER • October 2000 • www.activetradermag.com 8


TRADING Strategies

Anticipating BREAKOUTS
and beating SLIPPAGE
Trading breakouts is a tried-and-true

approach on all time frames. But intraday

and other short-term traders

can sometimes give up

precious points because

of slippage.

Here’s one trader’s take

on finding setups that allow

you to enter early and beat

the breakout crowd.

9 www.activetradermag.com • August 2000 • ACTIVE TRADER


BY STEVE WENDLANDT

O ne of the most important


aspects of short-term
stock trading is some-
thing you almost never
hear about: Slippage.
Slippage is the difference between
where you expect, or want, to be filled
more a particular level is tested, the
weaker it becomes.
In layman’s terms, if a stock continu-
ally prints or finds support or resistance
at a certain price, the odds are extremely
good that price level will be broken
shortly. That is invaluable information
who bought the stock around $50 will
either be stopped out or will wait for an
opportunity to breakeven on their
trades. The bottom line is that when sup-
port at 50 is penetrated it quickly turns
into significant resistance.
Here’s the question: If, because of
on a trade and where your order is actu- for any trader who uses breakouts as repeated tests of the support level, the
ally executed. If you don’t understand part of his or her strategy. odds are very good the 50 level will be
this concept, try to enter a market order Figure 1 is a five-minute chart of broken (and the broader market indices
with a browser-based online broker the CMGI. The stock bounced off support at support this view), why wait for the
first day of a hot IPO and see what hap- 50 six times (and who knows how many breakout? Doing so increases the odds of
pens. That’s slippage! Slippage can be prints actually occurred at that level). having to chase the market or missing the
caused by a number of factors: Poor exe- Every time a stock tests a support or trade. In this case, if you wait for the stock
cution by a broker, communication fail- resistance level, that level gets weaker to trade at 49 15⁄16 and then try to establish
ure or other technical problems, or fast and weaker, as if a hammer and chisel a short position, you’ll probably end up
market conditions. were chipping away at it. missing the trade waiting for an uptick.
While it’s true that we all try to keep Fortunately, most people view sup- Let’s look at a second example. In
our costs down to the bare minimum port levels as opportunities to go long, Figure 2, Netro Corp. (NTRO) was
without sacrificing service or technolo- while breakout traders view tests of sup- bouncing off the 82 1⁄2 level for about two
gy, slippage is probably the most over- port as fuel to propel an eventual break- weeks. The day it finally broke that sup-
looked and significant cost in trading. out. In this example, not only are traders port level (March 30, 2000) was a very
But through a little-known tendency, establishing new long positions with weak day in the broader market indices,
you can make slippage work for you their stops just below the support level which helped the stock to finally break
instead of bleeding you dry. In fact, if at 50, there are also many traders wait- down. A good opportunity to short
most of your trading techniques are ing to short the stock once it does break NTRO came at the prior day’s close
breakout related, you can use this trick down. Don’t forget that all the people when NTRO closed right at the support
on almost every trade you enter. But
first, let’s look at why it works.
FIGURE 1 CHISELING AT SUPPORT
One tick at a time Repeated tests of a support level increase the odds of a downside breakout.
Tom DeMark, a highly regarded trading A short position can be established in anticipation, with a stop just above
system developer who has worked with the most recent swing high to protect against an upside reversal.
such top traders as George Soros, Paul
CMGI (CMGI), 5-minute
Tudor Jones and Steve Cohen, wrote a 61
book (his second) called New Market 10:00 11:00 12:00 13:00 14:00 15:00 16:0010:00 11:00 12:00 13:00 14:00 15:00 16:00
Timing Techniques: Innovative Studies in 59
Market Rhythm and Price Exhaustion
(1997, John Wiley & Sons, New York). In Stop placed at most 57
it, he explained what probably is one of recent swing high (50 3⁄4)
55
the most significant discoveries in the
markets: the TD One-Tick, One-Time 53
Rule.
This rule states if a market makes a 51 1⁄8
CMGI repeatedly tests support 51
new high or low just once (a single print) at 50 in a weak market
and backs off from that point, that new 49
high or low should hold for a significant
period of time. In fact, most significant 47
highs and lows only print one time at the
extreme price.
350,500
It makes sense that the opposite also is
true: If a price prints more than once at a
certain high or low, then that high or low
will be broken in short order almost Source: CyberTrader by CyberCorp.
every time. From that, it follows the

ACTIVE TRADER • August 2000 • www.activetradermag.com 10


FIGURE 2 EARLY OPPORTUNITIES
A close at the low of the bar preceding the downside breakout, just
at the support level, offers an early entry opportunity for a short position. difficult to get short once a stock breaks
through support, if the trade is any
Netro Corp. (NTRO), daily good. You must either wait for an uptick
122
Jan. Feb. Mar. Apr. May (which may not happen) or offer it short
109 1⁄2
1
⁄16 higher than the inside bid (for Nasdaq
stocks). But if the stock is dropping like a
Stock tests support
in weak market. 97 rock, who is going to hit your offer?
Short trade entered The bottom line is that if you want to
at 82 9⁄16 841⁄2 trade a stock when the overall market is
trending in the direction of your poten-
72 tial trade, and the stock repeatedly tests
a support or resistance level, you should
591⁄2 enter before the breakout. Most times,
you even can avoid paying the spread
47
because the stock will be whipsawing
back and forth between the bid and offer.
341⁄2
If you wait until the stock breaks out you
25 11⁄16 are almost always forced to pay the
spread — if you can get it at all.
5,820,000
But, you may ask, what if the stock
never breaks out? Should you hold the
position until it does, or should you exit
Source: CyberTrader by CyberCorp. the position on the close? One approach
to reduce risk is to use the last swing low
level for the second day in a row. The market opened for trading on the day of or high as your initial stop-loss point. In
next morning NTRO gapped lower and the breakdown (although, there were the CMGI example, you could have
continued to drop dramatically. It would some upticks in the pre-market). placed an initial stop loss at 50 3⁄4 which
have been difficult to get short after the All breakout traders know it’s very was the last swing high on the five-
minute chart. With a stop in place, you
can simply wait for the breakout to mate-
FIGURE 3 GOING WITH THE MARKET rialize. The only reason not to hold the
position is if the overall market begins to
Pre-breakout entry should be confirmed by the broader market indices. In this
move counter to the trade (i.e., you’re
case, establishing a position in advance of a breakout above the trendline was
long, waiting for the breakout, and the
supported by strength in the S&P 500 and Nasdaq indices.
market begins to drop precipitously).
Warner Lambert (WLA), daily But you must use caution when enter-
130 ing breakout trades early; you never
Jan. Feb. Mar. Apr. May
125 9⁄64 want to enter a trade that is counter to
1231⁄2
the overall market momentum. For
Stock tests trendline 117 example, before entering the CMGI
resistance in strong trade on the short side, you should have
overall market. 1101⁄2 checked to make sure the Nasdaq and
Entered long at 122 1⁄4
(before the breakout). S&P 500 were both weak on the day and
104
trending lower. The weakness of these
971⁄2
indices would help pull the stock below
the support level.
91 Figure 3 shows one last example. On
May 25, Warner Lambert (WLA) opened
841⁄2 for trading at 121 1⁄2, just under the down
trendline of a nice triangle pattern. The
pre-opening call was for the Nasdaq and
8,434,900 S&P 500 to go higher that morning, and
they both began to rally from the open.
This created a setup to go long before
the actual breakout above the trendline.
Source: CyberTrader by CyberCorp. As soon as WLA began to move toward
the trendline, a buy order was entered at

11 www.activetradermag.com • August 2000 • ACTIVE TRADER


FIGURE 4 BREAKOUT PATTERNS

A sampling of the breakout patterns short-term traders can use on any time
frame. They provide well-defined support or resistance levels you can use to
anticipate breakouts.

Cup and handle breakout Trendline breakout

Spike and ledge breakout Triangle breakout

122 1⁄4, well before the 123 1⁄16 breakout breakouts to materialize on any of these
point. Not long after, the overall market patterns. Remember, slippage affects
strength helped pull WLA through the you whether or not you make a profit on
trendline; it continued to rally for the the trade. Most traders don’t even think
rest of the day. about the effect of slippage on their win-
Had you waited for WLA to print at ning trades; they only think about the
123 1⁄16, you would have been filled at a losers. And don’t forget about the trades
minimum of 13⁄16 worse than the early you missed completely because the
entry price. Those extra fractions add up stock just ripped through the support or
quickly. You can usually gain an extra 1⁄8 resistance level and you couldn’t even
(sometimes as much as a point) simply get a partial fill.
by realizing that support and resistance We tend to forget about those missed
almost always get broken. Try the fol- opportunities completely, but those are
lowing experiment: Multiply 50 percent usually the most potentially profitable
of all the shares you have traded over a trades because the stock is moving so
given time period by 1⁄8 and see what you forcefully. This approach will also help
come up with. That’s being conservative. you on the breakout trades that don’t
You can use this entry technique on materialize because you’ll have a better
any breakout-related trade in any time- entry price and may even be able to still
frame, including breakouts from daily garner a small profit or, at worst, scratch
and intraday cup-and-handle patterns, a trade from these false breakouts.
triangles, trendline breakouts and spike No approach is without risk, but in
and ledge patterns (see Figure 4). Very certain situations entering early can
rarely should you wait for the actual yield excellent trading results. 

ACTIVE TRADER • August 2000 • www.activetradermag.com 12


100-20 channel FIGURE 1 EQUITY CURVE
The long- and short-only equity curves, along with
breakout system the overall equity curve, are shown here. The long side
of the system substantially outperformed the short side
during the 10-year test period
System concept: This is a classic trend-following system that 190,000
buys when price moves above the highest high of the last x 180,000
days and sells when price falls below the lowest low of the last 170,000
160,000
y days. The number of days used to calculate the breakout level
150,000
is called the “channel length.” 140,000
Breakout systems are based on the logic that by making a 130,000
new price high (or low), a market is demonstrating it has the 120,000
momentum to establish a trend, and price will likely continue 110,000

Account balance ($)


in that direction. 100,000
90,000
In this test, one long channel length (100 days) was used for
80,000
entries, and a short channel length (20 days) was used for exits. 70,000
The exit strategy allows the system to follow large moves until 60,000
price makes a significant reversal. 50,000
We will also examine the results of using a range of channel 40,000
lengths and how a “walk-forward optimization” could 30,000
20,000
improve the results of the system for the most recent year.
10,000
0
Rules: 3/3/93 3/2/94 3/1/95 2/6/96 2/3/97 2/2/98 1/7/99 1/3/00 1/2/01 1/2/02 1/2/03
1. Enter long on the next bar at the highest 100-day high. Equity Cash Linear reg Long Short

2. Exit long on the next bar at the lowest 20-day low.


3. Enter short on the next bar at the lowest 100-day low. 4. Exit short on the next bar at the highest 20-day high.

Money management: Risk a maximum of 2 per-


FIGURE 2 SAMPLE TRADES
cent of total account equity per trade. The position
This short trade was triggered when price crossed below the 100-day low. size is based on the difference between the entry
The exit occurred when price crossed above the 20-bar high. The 100- and price and the initial stop level. Trade the number
20-day high/low channels are plotted as gray lines. of shares that would result in a 2-percent loss of
50.00 account equity if the stop level were hit.
Boeing (BA), daily
48.00
Starting equity: $100,000. Deduct $10 slippage and
46.00 commission per trade.
Short
44.00
Test data: The system was tested on the Active
42.00 Trader Standard Stock Portfolio, which contains
the following 18 stocks: Apple Computer (AAPL),
40.00 Boeing (BA), Citibank (C), Caterpillar (CAT),
38.00 Cisco (CSCO), Disney (DIS), General Motors
(GM), Hewlett Packard (HPQ), International
36.00 Business Machines (IBM), Intel (INTC),
34.00 International Paper (IP), JP Morgan Chase (JPM),
Coke (KO), Microsoft (MSFT), Sears (S), Starbucks
32.00 (SBUX), AT&T (T) and Wal-Mart (WMT).
Cover 30.00
Test period: January 1993 through February 2003.
10.00 M
Volume System results: The system’s performance was
5.00 M mediocre, at best: It returned only 12.61 percent
over 10 years, while buy and hold would have
August 2002 September October November returned more than 253 percent. Furthermore, the
Source for all figures: Wealth-Lab Inc. (www.wealth-lab.com)
system was exposed to the market nearly 75 per-

13 www.activetradermag.com • June 2003 • ACTIVE TRADER


cent of the time, which
means we are squeezing just TABLE 1 BEST PARAMETER FIGURE 3 DRAWDOWN CURVE
about as much performance VALUES FOR EACH The system was never able to overcome the drawdown
out of this system as possi- STOCK that began in mid-1995.
0%
ble, short of using margin or Symbol Long Short
period period -5%
some other form of leverage.
It is interesting to note, AAPL 70 16 -10%
however, that the long side BA 70 14
-15%
of the system performed C 130 26
much better than the short CAT 130 14 -20%

side. The net return for long CSCO 80 24 -25%


trades was 56 percent, with DIS 70 18
GM 80 18 -30%
only 38-percent market
exposure. Maximum draw- HPQ 90 24 -35%
IBM 90 18 3/3/93 3/3/94 3/1/95 2/9/96 2/3/97 2/2/98 1/8/99 1/3/00 1/2/01 1/2/02 1/2/03
down for the long side of the
INTC 70 16
system was only 18 percent, ket began to fall in the year 2000, and although this period did
IP 130 14
while buy and hold experi- produce a small profit, it was also accompanied by extreme
JPM 130 26
enced a devastating 66 per- KO 130 16 volatility.
cent maximum drawdown. MSFT 120 18
These results confirm S 70 14 System parameters: One way many traders attempt to
short trading in equities can SBUX 120 16 improve a system is to “optimize” its parameters (in this case,
be tricky. We measured the T 90 26 the number of days used to determine the channel lengths).
results of the short side of the WMT 110 14 This involves testing various parameter combinations to find a
system after the broad mar- range of values that result in the greatest profit over a given
period.
STRATEGY SUMMARY Although this technique can result in a system that shows
tremendous profit over a historical testing period, the odds that
Profitability Trade statistics you would have known to use those specific parameter values
Net profit ($): 12,608 No. trades: 330
Net profit (%): 12.61 Win/loss (%): 38.79
Exposure (%): 73.36 Avg. gain/loss (%): 0.09 PERIODIC RETURNS
Profit factor: 1.05 Avg. holding time: 34.09
Payoff ratio: 0.25 Avg. profit (winners): 12.67 Avg. Sharpe Best Worst Percentage Max. Max.
Recovery factor: 0.35 Avg. hold time (winners): 53.33 return ratio return return profitable consec. consec.
periods profitable unprofitable
Drawdown Avg. loss (losers) %: -7.88 Weekly 0.04% 0.15 11.49% -8.47% 49.42% 11 9
Max. DD (%): 35.29 Avg. hold time (losers): 21.91 Monthly 0.19% 0.15 13.53% -8.31% 50.83% 6 6
Longest flat days: 1,766 Max. consec. win/loss: 6/14 Quarterly 0.51% 0.15 22.09% -13.63% 48.78% 5 4
Annually 2.19% 0.17 33.91% -10.28% 50.00% 3 2
LEGEND: Net profit — profit at end of test period, less commission •
Exposure — the area of the equity curve exposed to long or short positions, as
LEGEND: Avg. return — the average percentage for the period • Sharpe ratio
opposed to cash • Profit factor — gross profit divided by gross loss • Payoff
— average return divided by standard deviation of returns (annualized) •
ratio — average profit of winning trades divided by average loss of losing
Best return — best return for the period • Worst return — worst return for
trades • Recovery factor — net profit divided by max. drawdown • Max DD
the period • % Profitable periods — the percentage of periods that were prof-
(%) — largest percentage decline in equity • Longest flat days — longest
itable • Max. consec. profitable — the largest number of consecutive prof-
period, in days, the system is between two equity highs • No. trades — num-
itable periods • Max. consec. unprofitable — the largest number of consec-
ber of trades generated by the system • Win/Loss (%) — the percentage of
utive unprofitable periods
trades that were profitable • Avg. profit — the average profit for all trades •
Avg. hold time — the average holding period for all trades • Avg. profit
(winners) — the average profit for winning trades • Avg. hold time (win- Trading System Lab strategies are tested on a portfolio basis (unless
ners) — the average holding time for winning trades • Avg. loss (losers) — otherwise noted) using Wealth-Lab Inc.’s testing platform.
the average loss for losing trades • Avg. hold time (losers) — the average If you have a system you’d like to see tested, please send the trad-
holding time for losing trades • Max. consec. win/loss — the maximum ing and money-management rules to editorial@activetradermag.com.
number of consecutive winning and losing trades

Disclaimer: The Trading System Lab is intended for educational purposes only to provide a perspective on different market concepts. It is not meant to recommend or
promote any trading system or approach. Traders are advised to do their own research and testing to determine the validity of a trading idea. Past performance does not
guarantee future results; historical testing may not reflect a system’s behavior in real-time trading.

ACTIVE TRADER • June 2003 • www.activetradermag.com 14


FIGURE 4 WALK-FORWARD OPTIMIZATION RESULTS
at the start of the period are about the same as picking the
After finding the optimal long and short channel lengths for each
winning lottery numbers for tomorrow. The parameters that
stock over the first nine years of historical data, we tested the
worked best in the past years are unlikely to be those that parameters on the most recent year of data. The system outper-
work best in the future. formed buy and hold (as well as the un-optimized parameters).
However, there are ways to use optimization effectively. One 120,000
technique is called “walk-forward optimization.” First, system 115,000
110,000
parameters are optimized on an initial (“sample”) data period. 105,000
100,000
Second, the best-performing parameters are used to execute the 95,000
system on a new, historical (“out-of-sample”) data period after 90,000
85,000
the sample period. This allows you to find out if the optimized

Account balance ($)


80,000
parameters would have improved the results going forward, 75,000
70,000
without cheating by using hindsight. 65,000
We performed a walk-forward optimization on the 100-20 60,000
55,000
channel breakout system by first optimizing the long and the 50,000
short channel periods for the first nine years of historical 45,000
40,000
price data. We then used the best-performing parameter val- 35,000
ues for each stock in the portfolio (see Table 1) and applied 30,000
25,000
them to the last year of historical price data. 20,000
15,000
Figure 4 is the equity curve for this optimized system. The 10,000
walk-forward optimized system lost 1.54 percent during the 5,000
0
one-year period, but buy and hold lost 30.57 percent. (The 3/1/02 4/3/02 5/7/02 6/12/02 7/22/02 8/28/02 10/7/02 11/15/02 1/2/03 2/6/03
system lost nearly 9 percent during this same year using the Equity Cash Linear reg Long Short Buy & holds

default parameter values of 100 and 20.) The walk-forward


optimization was effective in this case. with caution. The walk-forward technique described here can
The 100-20 channel breakout performs much better on the help you find more realistic optimized parameters that have a
long side than on the short side in stocks. Although it may be better chance of performing well in real trading.
possible to improve the system’s performance by optimizing
the channel periods for each stock, optimization must be used — Compiled by Dion Kurczek of Wealth-Lab Inc.

15 www.activetradermag.com • June 2003 • ACTIVE TRADER


FUTURES
Trading System Lab
FIGURE 1 EQUITY CURVE: 2 PERCENT MAXIMUM RISK
100-20 channel
The system equity curve with the 2-percent maximum loss setting has a
breakout system relatively stable uptrend.
System concept: The channel breakout is probably 220,000
one of the oldest trend-following systems around 210,000
(see the stock Trading System Lab on p. 46), and 200,000
one that has been especially popular in futures 190,000
markets over the years, for better or worse. 180,000
The system results published here are based on 170,000
a 100-day channel length for trade entries and a 20- 160,000
day channel length for exits. The channel lengths 150,000
are relatively long, because the system is intended 140,000

Account balance ($)


to catch long-term moves. 130,000
This system goes long and short. The stop levels 120,000
for both long trades and short trades play an 110,000
important role, because they are used to calculate 100,000
the position sizes in the different contracts. 90,000
80,000
Rules 70,000
1. Enter long on the next bar at the highest 60,000
100-day high. 50,000
2. Exit long on the next bar at the lowest 40,000
20-day low. 30,000
3. Enter short on the next bar at the lowest 20,000
100-day low. 10,000
4. Exit short on the next bar at the highest 0
20-day high. 3/25/93 3/1/94 2/1/95 1/4/96 1/2/97 1/2/98 1/4/99 1/3/00 1/2/01 1/2/02
(All trades are executed as stop orders.) Equity Cash Linear reg Long Short

Money management Source for all figures: Wealth-Lab Inc. (www.wealth-lab.com)


1. Risk a maximum of 2 percent of account equity
per trade. (Results will also be discussed for a 6-
FIGURE 2 EQUITY CURVE: 6 PERCENT MAXIMUM RISK percent maximum risk version of the system.)
2. To determine the position size (number of con-
The equity curve using a 6-percent maximum per-trade loss highlights tracts to trade), multiply the difference between the
large returns accompanied by high volatility and large drawdowns. entry price and the stop-loss price by the dollar
value of a one-point move in the contract, and
3,400,000 divide the result by the contract’s minimum margin.
3,200,000 For example, assume the contract being traded
3,000,000 has a point value of $250 and a $1,000 margin
requirement. Next, assume the initial entry buy
2,800,000
stop is at $100 (the value of the 100-day high) and
2,600,000 the initial stop-loss level is at 80 (the lowest 20-
2,400,000 day low). In this case, you would buy five [{(100 –
2,200,000 80)* $250}/$1000 = 5] contracts.
Account balance ($)

2,000,000
The $5,000 maximum loss this five-contract
trade represents should not be more than 2 per-
1,800,000
cent of the current portfolio equity. As a result,
1,600,000 unless the account equity is in excess of $250,000,
1,400,000 the system would not be able to take this position.
1,200,000
Starting equity: $100,000. Deduct $10 slip-
1,000,000
page/commission per trade.
800,000
600,000 Test data: The system was tested on the Active
400,000 Trader standard futures portfolio, which contains
the following 20 futures: DAX30 (AX), corn (C),
200,000
crude oil (CL), German bund (DT), Eurodollar
0 (ED), Euro Forex (FX), gold (GC), copper (HG),
Japanese yen (JY), coffee (KC), live cattle (LC),
3/25/93 3/1/94 2/2/95 2/1/96 1/8/97 1/2/98 1/4/99 1/3/00 1/2/01 1/2/02
lean hogs (LH), Nasdaq 100 (ND), natural gas
Equity Cash Linear reg Long Short (NG), soybeans (S), sugar (SB), silver (SI), S&P 500

ACTIVE TRADER • June 2003 • www.activetradermag.com 16


FIGURE 3 DRAWDOWN CURVE: 2 PERCENT MAXIMUM RISK
The maximum drawdown was about 18 percent.
0.00%
-2.00%
-4.00% cent.
The system results on the futures portfolio were fair-
-6.00%
ly good when the maximum risk was set to 2 percent
-8.00% per trade. The system returned an average profit of 8.42
-10.00% percent per year, with the largest losing year being -8.91
percent. The system’s market exposure was low — on
-12.00%
average, about 30 percent.
-14.00% Based on this information, the idea of increasing the
-16.00% risk and taking more contracts for each signal might
sound like a good idea, especially because there is still
-18.00%
3/25/93 3/1/94 2/1/95 1/9/96 1/2/97 1/2/98 1/4/99 1/3/00 1/2/01 1/2/02 plenty of margin available. Even though the system
reached an account value of more than $3 million (refer
(SP) and10 year T-Notes (TY). The test used Ratio FIGURE 4 DRAWDOWN CURVE: 6 PERCENT MAXIMUM RISK
Adjusted data from Pinnacle Data Corp
The drawdown increased both in depth and length in this version of the system.
0.00%
Test period: August 1993 to November 2002.
-5.00%
System results: Both the long and short sides of the sys-
tem were profitable, and the ratio of winning to losing -10.00%
trades was fairly balanced. The equity curve (Figure 1) -15.00%
using the 2-percent maximum loss setting shows a rela- -20.00%
tively smooth, steady uptrend. The 6-percent maximum -25.00%
loss version (Figure 2) posts a much larger gain, with
-30.00%
much higher volatility.
-35.00%
The position-sizing method keeps the system out of
many risky positions, although it resulted in no trade sig- -40.00%
nals in some markets because the risk was too high -45.00%
throughout the entire test period. -50.00%
To show the effect of the amount of risk taken, compare 3/25/93 3/1/94 2/1/95 1/9/96 1/2/97 1/2/98 1/4/99 1/3/00 1/2/01 1/2/02
the drawdown curves in Figures 3 and 4. Figure 3 is the
drawdown curve using a maximum risk of 2 percent. The maximum to Figure 2), the accompanying drawdown would have been near-
drawdown during this period was approximately 18 percent. Figure ly impossible to stomach. Exposure climbed near 70 percent, and
4 shows the result of increasing the maximum per-trade risk to 6 the longest wait between new equity highs was more than 750
percent. The effect is dramatic: The drawdown increased to 50 per- trading days.
The 100-20 channel breakout performed fairly well in this test. As
STRATEGY SUMMARY discussed in the stock Trading System Lab, you can experiment
with the system by optimizing the channel periods for each market.
Profitability Trade statistics — Compiled by Dion Kurczek of Wealth-Lab Inc.
Net profit ($): 99,997.48 No. trades: 292
Net profit (%): 100.00 Win/loss (%): 45.21
Exposure (%): 29.99 Avg. gain/loss (%): 0.73
Profit factor: 1.50 Avg. holding time: 37.60 PERIODIC RETURNS
Payoff ratio: 1.64 Avg. gain (winners) %: 6.22 Avg. Sharpe Best Worst Percentage Max. Max.
Recovery factor: 3.21 Avg. hold time (winners): 56.30 return ratio return return profitable consec. consec.
periods profitable unprofitable
Drawdown Avg. loss (losers) %: -3.80 Weekly 0.15% 0.64 7.29% -6.14% 52.07% 6 9
Max. DD (%): 19.59 Avg. hold time (losers): 22.17 Monthly 0.66% 0.61 12.32% -9.25% 55.08% 6 6
Longest flat days: 685 Max. consec. win/loss: 5/7 Quarterly 1.99% 0.55 24.09% -8.25% 40.00% 4 6

LEGEND: Net profit — profit at end of test period, less commission • Annually 8.58% 0.58 29.99% -8.91% 66.67% 3 2
Exposure — the area of the equity curve exposed to long or short positions, as
opposed to cash • Profit factor — gross profit divided by gross loss • Payoff LEGEND: Avg. return — the average percentage for the period • Sharpe
ratio — average profit of winning trades divided by average loss of losing ratio — average return divided by standard deviation of returns (annualized)
trades • Recovery factor — net profit divided by max. drawdown • Max DD • Best return — best return for the period • Worst return — worst return
(%) — largest percentage decline in equity • Longest flat days — longest for the period • % Profitable periods — the percentage of periods that were
period, in days, the system is between two equity highs • No. trades — num- profitable • Max. consec. profitable — the largest number of consecutive
ber of trades generated by the system • Win/Loss (%) — the percentage of profitable periods • Max. consec. unprofitable — the largest number of
trades that were profitable • Avg. gain — the average profit for all trades • consecutive unprofitable periods
Avg. hold time — the average holding period for all trades • Avg. gain
Trading System Lab strategies are tested on a portfolio basis (unless
(winners) — the average profit for winning trades • Avg. hold time (win-
ners) — the average holding time for winning trades • Avg. loss (losers) — otherwise noted) using Wealth-Lab Inc.’s testing platform.
the average loss for losing trades • Avg. hold time (losers) — the average If you have a system you’d like to see tested, please send the trad-
holding time for losing trades • Max. consec. win/loss — the maximum ing and money-management rules to editorial@activetradermag.com.
number of consecutive winning and losing trades
Disclaimer: The Trading System Lab is intended for educational purposes only to provide a perspective on different market concepts. It is not meant to recommend or
promote any trading system or approach. Traders are advised to do their own research and testing to determine the validity of a trading idea. Past performance does not
guarantee future results; historical testing may not reflect a system’s behavior in real-time trading.

17 www.activetradermag.com • June 2003 • ACTIVE TRADER


FUTURES
Trading System Lab
60-minute breakout system
Market: Futures (indices).
FIGURE 1 EQUITY CURVE
System concept: This is an intraday system The system produced a modest profit, with long trades outperforming shorts.
that trades on breakouts of the range estab-
lished in the first hour of trading. For a detailed 24,000
explanation of the strategy please read the 22,000
stock Trading System Lab on p. 50.
The intention was to see how the system per- 20,000
formed on stock index futures as opposed to 18,000
individual stocks. In this test the S&P 500 (SPY)
16,000
and Nasdaq 100 (QQQ) index-tracking stocks

Account balance ($)


were used as proxies for the S&P 500 and 14,000
Nasdaq 100 futures.
12,000

Entry rules: 10,000


Long trades: Buy if the closing price of the
8,000
third 30-minute bar is above the high of the
first 60 minutes of the day. 6,000

4,000
Short trades: Sell short if the closing price of
the third 30-minute bar is below the low of the 2,000
first 60 minutes of the day. 0
10/15/01 1/8/02 4/1/02 6/24/02 9/23/02 12/30/02 4/2/03 6/26/03 9/25/03
Exit: Exit all positions on signals in the
opposite direction or at the end of the day. Equity Cash Long Short Buy & hold

Money management: To equalize the weight of Source for all figures: Wealth-Lab Inc. (www.wealth-lab.com)

both markets, 49 percent of the current portfo-


lio capital is allocated for every trade. For FIGURE 2 DRAWDOWN CURVE
example, if the total equity moves up to $22,000
The largest drawdown occurred early in the test period. The system’s
and our strategy generates a new signal, we biggest string of losing trades was seven.
would invest $10,780 for the next signal. We
use 49 percent to give us some leeway for com- 0%
-1%
mission. Please keep in mind that we use the
-2%
portfolio result and not the individual result.
-3%
This is very important and should always be -4%
used since only this method reflects what you -5%
would actually experience later in your trad- -6%
ing. -7%
-8%
Starting equity: $20,000 (nominal). Deduct -9%
$0.01 per share slippage and commissions. -10%
-11%
-12%
Test period: October 2001 until October 2003.
-13%

Test data: SPY and QQQ. The SPY is designed 10/15/01 1/2/02 3/14/02 5/30/02 8/9/02 10/28/02 1/21/03 4/2/03 6/13/03 8/29/03
to trade at one-tenth the level of the S&P 500;
the QQQ is designed to trade at one-fortieth of the Nasdaq 100. close all positions not triggered by an opposite signal at the
Like futures, the uptick rule to enter short positions does not close of the day.
apply to these instruments. QQQ and SPY can be traded intra-
day but have the advantage that no rollover occurs every three Test results: The results for the two years are very encourag-
months. ing: a profit of 19.88 percent on the starting capital in two
We downloaded more than two years of 30-minute bars from years, compared to an unchanged result for the combined
the QCharts historical intraday database for SPY and QQQ. equities of the two indices (see Figure 1).
There are a few interesting things to note. For the first hour The system generated its largest drawdown (-13.52 per-
range we take the prices from 9:30 a.m. to 10:30 a.m. and for the cent) on Feb. 21, 2002 (see Figure 2). Buy and hold’s largest
closing time we use 4:15 p.m. This is important because we will drawdown (on Oct. 9, 2002) was -44.87 percent.

18 www.activetradermag.com • January 2004 • ACTIVE TRADER


FIGURE 3 SAMPLE TRADES
The average hold time for both winning and losing trades was around seven days.

On the downside, the average profit 34.10


Nasdaq 100 index-tracking stock (QQQ), 30-minute
per trade was just 0.05 percent, or $4.46.
This may be too little to really trade the 34.00
system. Looking at the statistics it is inter-
esting to note that out of the 892 trades 33.90
over the last two years, only 102 were
stopped out by the opposite signal while 33.80
the rest stayed with the initial direction. It
Sell
seems that in most cases, once the market 33.70
begins an intraday trend, it continues in
that direction throughout the day. 33.60
Figure 3 shows a short trade on Sept.
10, 2003, that was exited at the close of the 33.50
day. On the following day it appeared the Buy
market was continuing its down move. 33.40
We received a short signal but got
33.30
stopped out after the market bounced
Sell
back.
33.20

Bottom line: There is a big difference


33.10
between indices and stocks in regard to
this system. Individual stocks tend to be Buy
33.00
much more volatile than an index; also,
with an index you hardly see large 9/10/03 9/11/03
overnight gaps. This might be one reason
the 60-minute breakout system performs much better on the Because of the low average profit per trade, the system
ETFs than it does on the individual stocks. requires more fine tuning. Nevertheless, the ratio of trades
that kept their original position for the whole day makes this
strategy worthy of further investigation.
STRATEGY SUMMARY
— Volker Knapp of Wealth-Lab Inc.
Profitability Trade statistics
Net profit ($): 3,976.80 No. trades: 892
Net profit (%): 19.88 Win/loss (%): 54.60
Exposure (%): 44.95 Avg. gain/loss (%): 0.05 PERIODIC RETURNS
Profit factor: 1.11 Avg. hold time: 7.21
Avg. Sharpe Best Worst Percentage Max. Max.
Payoff ratio: 0.93 Avg. profit (winners) %: 0.80 return ratio return return profitable consec. consec.
Recovery factor: 1.38 Avg. hold time (winners): 7.35 periods profitable unprofitable
Drawdown Avg. loss (losers) %: -0.86 Weekly 0.19% 0.80 7.31% -3.36% 53.85% 7 8
Max. DD (%): -13.52 Avg. hold time (losers): 7.04 Monthly 0.77% 0.88 6.97% -3.17% 52.00% 2 4
Longest flat days: 1,742 Max. consec. win/loss: 10/7 Quarterly 2.07% 1.39 6.19% -2.19% 66.67% 6 2
LEGEND: Net profit — Profit at end of test period, less commission •
Exposure — The area of the equity curve exposed to long or short positions, LEGEND: Avg. return — The average percentage for the period • Sharpe
as opposed to cash • Profit factor — Gross profit divided by gross loss • ratio — Average return divided by standard deviation of returns (annual-
Payoff ratio — Average profit of winning trades divided by average loss of los- ized) • Best return — Best return for the period • Worst return — Worst
ing trades • Recovery factor — Net profit divided by max. drawdown • return for the period • Percentage profitable periods — The percentage of
Max. DD (%) — Largest percentage decline in equity • Longest flat days — periods that were profitable • Max. consec. profitable — The largest num-
Longest period, in days, the system is between two equity highs • No. trades ber of consecutive profitable periods • Max. consec. unprofitable — The
— Number of trades generated by the system • Win/Loss (%) — The per- largest number of consecutive unprofitable periods
centage of trades that were profitable • Avg. gain — The average profit for all
trades • Avg. hold time — The average holding period for all trades • Avg. Trading System Lab strategies are tested on a portfolio basis (unless
gain (winners) — The average profit for winning trades • Avg. hold time otherwise noted) using Wealth-Lab Inc.’s testing platform.
(winners) — The average holding time for winning trades • Avg. loss (los- If you have a system you’d like to see tested, please send the trad-
ers) — The average loss for losing trades • Avg. hold time (losers) — The ing and money-management rules to editorial@activetradermag.com.
average holding time for losing trades • Max. consec. win/loss — The max-
imum number of consecutive winning and losing trades
Disclaimer: The Trading System Lab is intended for educational purposes only to provide a perspective on different market concepts. It is not meant to recommend or
promote any trading system or approach. Traders are advised to do their own research and testing to determine the validity of a trading idea. Past performance does not
guarantee future results; historical testing may not reflect a system’s behavior in real-time trading.

ACTIVE TRADER • January 2004 • www.activetradermag.com 19


Four-percent breakout system
FIGURE 1 SAMPLE TRADES
Market: Nasdaq 100 index-tracking stock (QQQ).
March and April 2003 were very active months for the four-percent
System concept: This system is from the book Trade breakout system. There was one large winner, two mid-size winners
like a Hedge Fund (John Wiley & Sons, 2004) by and one large losing trade.
James Altucher. Nasdaq 100 index-tracking stock (QQQ), daily 27.40
Both day traders and hedge-fund managers love Sell Sell 27.20
to “fade” (i.e., trade in the opposite direction of) 27.00
sharp intraday moves. The thought behind this 26.80
Sell 26.60
type of trading is price is unlikely to go much high- 26.40
er after an extreme up move. Traders take short Sell
26.20
positions anticipating a reversal. 26.00
In some situations, however, the market ignores Buy Buy 25.80
the contrarians and continues to rise. Traders who 25.60
25.40
fade the up move must cover their short positions, 25.20
which leads to panic buying and further upward Buy 25.00
momentum. 24.80
The four-percent breakout system is an attempt 24.60
24.40
to quantify and profit from this market scenario.
Buy 24.20
The system goes long when price rises four percent 24.00
from the previous close — in this case, assumed to 23.80
be the point at which short sellers must concede 23.60
they were wrong and cover their positions, driving 23.40
prices even higher during the trading day. The sys- Volume
100M
tem is long only; no short trades are made.
50M
Rules: April 2003
Entry — Buy today if price gains four percent Source for all figures: Wealth-Lab Inc. (www.wealth-lab.com)
from the previous trading day’s closing price.
Exit — Exit on the open of the next trading day.
FIGURE 2 EQUITY CURVE
Figure 1 shows sample trades in QQQ from March The equity grew steadily from 2000 through 2002, but the system
and April 2003. has been stagnating since mid-2002.
250,000
Risk control and money management: This system 240,000
tests only one market, and enters only one position at 230,000
a time, so 100 percent of equity should be tied up on 220,000
210,000
each trade. 200,000
190,000
Starting equity: $100,000. Deduct $20 per round-turn 180,000
170,000
trade for slippage and commissions. 160,000
Account balance ($)

150,000
Test data: The system was initially tested only on the 140,000
130,000
QQQ. It was also tested on the Active Trader Standard 120,000
Stock Portfolio, which contains the following 18 110,000
stocks: Apple Computers (AAPL), Boeing (BA), 100,000
90,000
Citigroup (C), Caterpillar (CAT), Cisco Systems 80,000
(CSCO), Disney (DIS), General Motors (GM), Hewlett- 70,000
Packard (HPQ), International Business Machines 60,000
50,000
(IBM), Intel (INTC), International Paper (IP), J.P. 40,000
Morgan Chase (JPM), Coca-Cola (KO), Microsoft 30,000
(MSFT), Sears (S), Starbucks (SBUX), AT&T (T) and 20,000
10,000
Wal-Mart (WMT). 0
3/10/99 9/1/99 3/1/00 9/1/00 3/1/01 9/4/01 3/8/02 9/5/02 3/6/03 9/3/03 3/3/04
Test period: March 1999 through June 2004 for the
Equity Cash
QQQ test; July 1994 to June 2004 for the Active Trader
portfolio.

20 www.activetradermag.com • September 2004 • ACTIVE TRADER


FIGURE 3 DRAWDOWN CURVE
The drawdown phase from mid-2002 to the present dominates the
drawdown curve.
0%
-1%
-2%
-3%

Drawdown
-4%
Test results: Figure 1 shows the first trade was a suc- -5%
cess. Price rose 40 cents after entry and the market -6%
made a small gap open the following day for a two- -7%
percent profit. The next two trades, which occurred -8%
only a few days later, were not as successful but -9%
nonetheless booked modest profit. -10%
However, the next trade wiped out the previous -11%
profit and then some. Price gapped up at the market 3/10/99 9/1/99 3/2/00 9/1/00 3/5/01 9/4/01 3/8/02 9/6/02 3/7/03 9/5/03 3/5/04
open, beyond the four-percent threshold, and the
entry order was filled (this particular trade would
have probably been subject to negative slippage because of the the designers knew of previous QQQ price movement), it was
volatility at the open). tested on other markets in an attempt to determine its validity.
Price then suddenly reversed, and the result was a large loss Our starting equity for the Active Trader portfolio test was
upon the exit the following day. The fact that the initial losing also $100,000, although only 10 percent of equity was commit-
trade occurred on a day when prices gapped above the entry ted per trade.
level on the open suggests the system might benefit from a fil- This equity curve (Figure 4, p. 60) shows fairly steady
ter that ignores the signal if price opens with a greater than growth from the beginning of the test period through mid-
four-percent gain. 2002. From that point, there is a slight decline in capital and a
The equity curve (Figure 2) provides a better indication of general stagnation as fewer trades take place.
the system’s overall performance. After a small loss in 1999, This equity curve mirrors the QQQ equity curve. However,
profits began in early 2000 and lasted until mid- to late 2002. the fact that the system was profitable on a portfolio of stocks
The drawdown curve (Figure 3) confirms this, as the 12-per- (8.95 percent annualized gain) and not just one stock is evi-
cent drawdown began in late 2002. The system is more or less dence the system is based on a valid core assumption.
flat from April 2003 forward. The only trade after that was in
July 2003, resulting in a loss of 0.08 percent. System variation: James Altucher publishes a variation of the
system that adds one additional entry rule: Price must be down
Portfolio test results: While it is still too early to tell if this sys- two percent on the day before entering a trade. This rule is
tem is worth trading on the QQQ (because it’s possible the sys- intended to avoid entering when a price move is nearly
tem was subconsciously designed to take advantage of what exhausted, and allows the system to capture solid rebound

STRATEGY SUMMARY PERIODIC RETURNS


Profitability Trade statistics
Avg. Sharpe Best Worst Percentage Max. Max.
Net profit ($): 121,023 No. trades: 105 return ratio return return profitable consec. consec.
Net profit (%): 121.09 Win/loss (%): 64.76 periods* profitable unprofitable
Exposure (%): 7.81 Avg. trade (%): 0.80 Weekly 0.30% 1.18 11.89% -9.23% 22.10% 4 63
Profit factor: 1.82 Avg. winner (%): 2.38 Monthly 1.30% 1.35 13.34% -6.40% 50.00% 10 15
Payoff ratio: 1.13 Avg. loser (%): -7.55
Quarterly 3.93% 1.03 25.74% -6.98% 59.09% 8 5
Recovery factor: 4.05 Avg. hold time: 1.00
Annually 16.71% 0.56 75.51% -2.52% 66.67% 4 2
Drawdown ($): 29,900 Avg. hold time (winners): 1.00
*The system remains flat much of the time. A flat period is considered unprofitable
Max. DD (%): -11.93 Avg. hold time (losers): 1.00 for purposes of this report.
Longest flat days: 420 Max. consec. win/loss: 11/5

LEGEND: Net profit — Profit at end of test period, less commission • Exposure LEGEND: Avg. return — The average percentage for the period • Sharpe ratio —
— The area of the equity curve exposed to long or short positions, as opposed to cash Average return divided by standard deviation of returns (annualized) • Best return
• Profit factor — Gross profit divided by gross loss • Payoff ratio — Average — Best return for the period • Worst return — Worst return for the period •
profit of winning trades divided by average loss of losing trades • Recovery factor Percentage profitable periods — The percentage of periods that were profitable •
— Net profit divided by max. drawdown • Max. DD (%) — Largest percentage Max. consec. profitable — The largest number of consecutive profitable periods •
decline in equity • Longest flat days — Longest period, in days, the system is Max. consec. unprofitable — The largest number of consecutive unprofitable periods
between two equity highs • No. trades — Number of trades generated by the sys-
tem • Win/Loss (%) — the percentage of trades that were profitable • Avg. trade
— The average profit/loss for all trades • Avg. winner — The average profit for Trading System Lab strategies are tested on a portfolio basis (unless
winning trades • Avg. loser — The average loss for losing trades • Avg. hold time otherwise noted) using Wealth-Lab Inc.’s testing platform.
— The average holding period for all trades •Avg. hold time (winners) — The
If you have a system you’d like to see tested, please send the
average holding time for winning trades • Avg. hold time (losers) — The aver-
age holding time for losing trades • Max. consec. win/loss — The maximum trading and money-management rules to editorial@activetradermag.com.
number of consecutive winning and losing trades
Disclaimer: The Trading System Lab is intended for educational purposes only to provide a perspective on different market concepts. It is not meant to recommend or
promote any trading system or approach. Traders are advised to do their own research and testing to determine the validity of a trading idea. Past performance does not
guarantee future results; historical testing may not reflect a system’s behavior in real-time trading.

ACTIVE TRADER • September 2004 • www.activetradermag.com 21


moves. This generally increases the efficiency FIGURE 4 EQUITY CURVE: ACTIVE TRADER PORTFOLIO
of the system while reducing the number of
actual trades. The upper equity curve shows the results of the four-percent breakout
The bottom equity curve in Figure 4 shows system on the Active Trader standard stock portfolio using 10 percent of
the results of the system variation on the equity per trade. The lower equity curve is a system variation that enters
Active Trader portfolio. Since the new filter after a down move and uses 25 percent of equity per trade.
reduces the number of trades, the position A
200,000
size was changed to 25 percent for each trade.
The shape of the equity curve is similar to the 150,000
previous run, but actual profit is higher and
100,000
the system does not enter as many losing

Account balance ($)


trades during the stagnation period of mid- 50,000
2002 to present.
0
250,000
Bottom line: The four-percent breakout sys- B
tem could not be much simpler. Simpler sys- 200,000
tems are often the most effective, and this 150,000
one is no exception. However, there needs to 100,000
be sufficient “post-publication” data to pro-
vide a reliable test for the QQQs. 50,000
0
This system is from James Altucher’s Trade 7/15/94 7/31/95 6/7/96 6/2/97 6/1/98 5/6/99 5/1/00 5/1/01 5/1/02 5/1/03 4/5/04
like a Hedge Fund. Equity Cash
— Compiled by Volker Knapp of Wealth-Lab

22 www.activetradermag.com • September 2004 • ACTIVE TRADER


TRADING Strategies

BROADENING PATTERNS:
Clues to breakout direction
A partial rise or decline can predict the direction of a breakout.
Learn to use these signals to increase profits when trading broadening patterns.

BY THOMAS N. BULKOWSKI

T rying to determine when a


breakout will occur in
broadening chart patterns,
which are expanding rather
than contracting price formations, can be
difficult. However, partial rises (PRs) or
partial declines (PDs) can improve the
odds of making a correct decision.
These signals predict immediate
breakouts and indicate their direction,
too, allowing you to increase your prof-
its and reduce your losses. However,
because a PR or PD often slows overall
momentum, the size of the eventual
breakout is not as large as when a PR or
PD does not appear.

Broadening tops and bottoms


Figure 1 shows two broad-
FIGURE 1 PARTIAL RISE ening bottom patterns.
These are different from
The July broadening bottom pattern appeared midway through the down move. A partial broadening tops because
rise accurately signaled a downside breakout from the pattern. price enters the pattern from
the top. In both patterns,
Milacron Inc. (MZ), daily
34 price touches each trendline
32 at least two times and
swings in a progressively
30
29 wider range. That is, the
Partial rise 28 minor highs get higher and
27
the minor lows get lower.
26
25 The July pattern shows a
34 PR, which occurs after the
12 24
23 pattern is established — that
22 is, there were at least two
21 touches of each trendline
20 before the PR. Price makes a
19 partial rise when it leaves the
18 bottom trendline and works
17 its way higher but fails to
touch or come too close to
16
the top trendline before turn-
15 ing away.
14 How close is “close”? Use
the figures in this article and
13 your common sense as
guides. For example, the
12
1998 Mar. Apr. May June July Aug. Sept. Oct. Nov. Dec. July broadening bottom has
Source: Proprietary software (Thomas Bulkowski) three top trendline “touch-

ACTIVE TRADER • April 2004 • www.activetradermag.com 23


TABLE 1 PARTIAL RISES AND DECLINES: SUCCESS RATES even better — 80 percent (see Table 1, top
The percentages reflect how often partial rises and partial declines predicted right, for more statistics).
breakout direction. Notice how the July pattern is mid-
way between the price at the start of the
Chart pattern Partial decline Partial rise downtrend (around 32) and its low
(around 14). The middle of the pattern is
Broadening bottom 80% 67% around 23, the center of the 32-14 range.
Broadening top 65% 86% Although broadening patterns some-
times act as “half-staff patterns” that
Broadening wedge, ascending Not measured 84%
form in the middle of moves, broaden-
Broadening wedge, descending 76% Not measured ing bottoms usually function as reversals
in a downtrend, not as continuation pat-
Right-angled and ascending Not measured Not measured
terns within those trends, as they do in
Right-angled and descending 78% 58% Figure 1.
Figure 2 includes two broadening
tops with PDs. In a partial decline, price
es,” not two or four: The second minor 500 stocks from mid-1991 to mid-1996, a leaves the top trendline and descends
high (point 2) comes close enough to call bull market, showed that a PR correctly but does not come close to or touch the
it a touch, but the third (point 3) does predicted a downward breakout 67 per- bottom trendline. An upward breakout
not. cent of the time. The accuracy rate of usually follows immediately. Again, the
Analysis of 77 broadening bottoms on PDs predicting upside breakouts was broadening top pattern must touch each
trendline at least two times
before a PD signal can occur.
FIGURE 2 BROADENING TOPS Table 1 shows PDs in
Both of these broadening tops included partial declines, which predicts an upward break- broadening tops correctly
out of the pattern. predicted an upward break-
out 65 percent of the time,
Newport Corporation (NEWP), daily while partial rises were 86-
percent accurate in predict-
7 ing downside breakouts.
In a larger combined
6 study of broadening tops
and bottoms, PDs worked 77
percent of the time. When a
Partial decline 5 PD occurred, the post-break-
out up move was 32 percent;
without a PD, the rise meas-
ured 36 percent. Thus, the
4 PD affected momentum by
reducing the eventual rally.
PDs not resulting in break-
Partial decline outs occurred just nine per-
3 cent of the time, which
means false signals are com-
paratively rare.
For PRs, the post-breakout
decline measured 15 percent;
without a PR, the declines
averaged 17 percent, indicat-
2 ing a partial rise steals ener-
1997 June July Aug. Sept. Oct. Nov. Dec. 1998 Feb. Mar. Apr. May June
gy from the resulting down
Source: Proprietary software (Thomas Bulkowski)
move.

24 www.activetradermag.com • April 2004 • ACTIVE TRADER


FIGURE 3 TRENDLINE TOUCHES
Look for a partial rise or decline only after price touches each trendline of the broaden-
ing pattern at least twice. Here, a partial rise formed in this ascending, right-angled
broadening formation.
Tommy Hilfiger (TOM), daily
37
Partial rise 35
33
False breakout signals for
PRs (i.e., when a partial rise 31
occurred inside a broadening 30
29
top pattern after price 28
1 2 3
touched each trendline twice 27
without triggering a break- Not a partial decline 26
out) occurred just 11 percent 25
24
of the time in the 350 pat-
23
terns examined.
22
A broadening top usually
21
acts as a continuation pattern
20
within the prevailing price
trend, as shown in Figure 2. 19
18
Right-angled broadening 17
formations
16
Figure 3 shows a “right-
angled,” ascending broaden- 15
ing formation. The top trend-
14
line slopes upward (ascends) 1998 Feb. Mar. Apr. May June July Aug. Sept. Oct.
and the bottom trendline is Source: Proprietary software (Thomas Bulkowski)
horizontal or nearly so. Like
other broadening patterns,
the breakout can occur in FIGURE 4 REVERSAL PATTERN
any direction, but this pat- Although broadening formations are often continuation patterns, descending, right-angled
tern usually reverses the broadening formations like the one shown here usually act as reversal patterns.
trend. The figure shows this,
as prices rise into the pattern CDI Corp. (CDI), daily
Partial rise 26
and exit out the bottom.
After two touches of each 25
trendline occur, look for a 24
partial rise or decline. The 23
late-May decline in Figure 3
22
does not show a partial
decline. Why? Because the 21
pattern at that point did not 20
have at least two minor 19
touches of each trendline.
Price touches at point 1 but it 18
is not a minor high or low, so
17
it does not count as a touch.
Point 2 is valid, as is point 3. 16
Only after price touches
point 3 can you draw the 15
horizontal trendline. By that
time, the three touches on 14
the top connect an up-slop-
ing trendline. The partial rise 13
that follows correctly pre-
dicts a downward breakout. 12
1994 Nov. Dec. 1995 Feb. Mar. Apr. May June July Aug. Sept.
Figure 4 shows a descend-
Source: Proprietary software (Thomas Bulkowski)
ing right-angled broadening

ACTIVE TRADER • April 2004 • www.activetradermag.com 25


FIGURE 5 DESCENDING BROADENING WEDGE
The August pattern did not produce a valid partial decline because price must drop from
the top trendline and curl around. Here, price rose from the bottom trendline. The first
partial decline in the October pattern fails to predict an immediate upward breakout, but pattern. The top trendline is
is correct in the longer term. horizontal and the bottom
one slopes down. Price
Transocean Inc. (RIG), daily touches the bottom trend-
52
line, bounces up but does not
50
48
come close to or touch the
46 top trendline before retrac-
44 ing its gains. This PR predict-
42
ed a downward breakout. As
is the case with this example,
40
the descending, right-angled
38
broadening pattern usually
36 acts as a price reversal. In the
34 descending pattern, partial
32 rises worked just 58 percent
31 of the time and partial
30 declines worked 78 percent
29
28 of the time in descending,
Not a 27 right-angled broadening pat-
partial decline 26 terns.
25
24
Broadening wedges
23 Figure 5 shows a descending
Partial declines 22
broadening wedge, which
21
consists of two down-slop-
20 ing trendlines (think of a
1999 June July Aug. Sept. Oct. Nov. Dec. 2000 Feb. Mar.
downward-tilting mega-
Source: Proprietary software (Thomas Bulkowski)
phone). The rules for wedges
are the same as other broad-
FIGURE 6 ASCENDING BROADENING WEDGE ening patterns: There must
After the pattern is established, a partial decline fails to correctly predict an upward be at least two minor high
breakout. Later, a partial rise precedes a downside breakout. touches of the top trendline
and at least two minor low
WPS Resources Corp. (WPS), daily touches of the bottom trend-
32
line. Only then is the pattern
31 valid and only then should
30 you look for a partial rise or
Partial rise decline.
29 The pattern usually acts as
28
a continuation, rather than a
reversal, of the prevailing
27 price trend. However, the
two wedges shown in Figure
26 5 are reversal patterns. In the
August pattern, prices
25
climbed into the pattern and
broke out to the upside, but
24
the overall trend (except for
23 a few days after the break-
out) was downward after the
Failed partial decline 22
pattern. Prices in the October
wedge were trending down-
ward into the pattern and
21
exited out its top. The trend
after the pattern ends is pre-
20 dominantly upward.
1999 Aug. Sept. Oct. Nov. Dec. 2002 Feb. Mar. Apr. May June
In the August pattern
Source: Proprietary software (Thomas Bulkowski)
example, the slight dip in

26 www.activetradermag.com • April 2004 • ACTIVE TRADER


early September was not a partial
decline. Price in a PD must start from the
top trendline, bow downward (without
coming close to or touching the bottom Additional research
trendline) and rejoin the top trendline. In
this case, price leaves the bottom trend- Books by Thomas Bulkowski:
line, not the top one.
In the October pattern, the first partial Encyclopedia of Chart Patterns (John Wiley & Sons, 2000)
decline is a failure because price does Trading Classic Chart Patterns (John Wiley & Sons, 2002)
not breakout upward immediately after
touching the top trendline. Instead, price
drops down again and finally shoots out Active Trader articles:
the top. “Technicals meet fundamentals in the earnings flag,” February 2004, p. 30
A partial decline correctly predicts an
upward breakout 76 percent of the time. “A different breed of scallop,” January 2004, p. 32
Not enough samples were found for par-
tial rises in descending broadening “The three rising valleys pattern,” December 2003, p. 28
wedges.
Figure 6 (p. 28) shows an ascending “Pipe bottom reversals,” November 2003, p. 28
broadening wedge. Both trendlines
slope upward and minor highs and
“Grabbing the bull by the horns,” September 2003, p. 46
minor lows touch each trendline at least
twice. The January partial decline failed
“Head-and-shoulders bottoms: More than meets the eye,” August 2003, p. 32
because price did not break out to the
upside — it touched the top trendline,
then reversed. “The high-low game,” July 2003, p. 28
The partial rise does better when it
leaves the bottom trendline, bounces up “Tom Bulkowski’s scientific approach,” September 2002, p. 32
and then plunges through the bottom
trendline. A PR correctly predicts a
downward breakout 84 percent of the
time.

ACTIVE TRADER • April 2004 • www.activetradermag.com 27


TRADING Strategies

HIGH, TIGHT FLAG


helps squeeze out profits
This bullish formation boasts excellent
post-breakout performance and a low failure rate —
exactly the type of pattern traders should look
for in bull markets.

BY THOMAS N. BULKOWSKI

A
is not over.
high, tight flag (HTF) is a
consolidation pattern that
forms after a stock’s price
doubles. When price breaks
out above the pattern, it signals the rise

Figure 1 shows an example of an HTF


moves following HTFs show how such
an approach can work.

Flag criteria
What should you look for when select-
ing HTFs? That depends on whom you
ask. William O’Neil, who popularized
that formed in January-February 2000. the pattern, has several selection criteria
The uptrend started in October at a low (see “Additional reading,” p. 33). He has
of 5.50 and reached a high of 11.35 at the written the rally preceding the pattern
HTF’s starting point — a doubling of should measure 100 to 120 percent and
price in less than a month. take less than two months; the flag
Although many HTFs have irregular should move sideways for three to five
shapes, you can usually draw a trendline weeks. Finally, the flag should retrace no
along the top of the pattern to signal a more than 20 percent of the preceding
breakout. In this example, parallel trend- rally.
lines mark the flag’s upper and lower Applying these rules to 252 patterns
boundaries. Volume slopes downward found in price data of approximately 500
over the course of the flag, as it did in 90 stocks between mid-1991 and early 2004
percent of HTFs in a recent study. filtered out all of them! (An earlier study
The basic HTF trade strategy is to buy found only six of 81 patterns met his crite-
at the close of the day after price breaks ria; these patterns did, however, produce
out above the pattern’s upper trendline. average gains of 69 percent.) The flag
In Figure 1, the stock rallied 52 percent shown in Figure 1 actually does not meet
from the closing price the day after price O’Neil’s criteria because it retraced 52
pierced the HTF’s upper trendline to the percent of the prior rise (most flags failed
ultimate high. O’Neil’s filter because they retraced more
Although it is sometimes difficult to than 20 percent) and the flag duration
buy high and sell higher, the price lasted more than seven weeks .

28 www.activetradermag.com • December 2004 • ACTIVE TRADER


FIGURE 1 A LARGE HIGH, TIGHT FLAG

This pattern does not meet William O’Neil’s HTF criteria, but the post-breakout rise was
52 percent. Such a well-defined flag shape is unusual.

Alkermes (ALKS), daily


17
Ultimate high 16
Another study that used 15
different selection criteria 14
for HTFs also showed an 13
average post-breakout gain 12
of 69 percent. The criteria for
11
this study was simply a near
doubling (a rise of 90 per- 10
cent or more in less than two 9
High, tight flag
months) of the stock price,
which is easy to find by 8
looking for stocks that have
moved up sharply in less 7
than two months, then
searching for a nearby con- 6
solidation region. The study
Trend start
placed no limit on flag
length, although HTFs equal 5
to or shorter than the 14-day
median length performed
better (71 percent) than 4
1998 Aug. Sept. Oct. Nov. Dec. 1999 Feb. Mar. Apr. May June
those that were longer (66
percent). The study also Source: Proprietary software (Thomas Bulkowski)
ignored the size of the
retracement. FIGURE 2 TWO HTFS
The first flag looks like a descending, broadening wedge and the second like a falling
HTF examples wedge. Both HTFs show good gains after the breakout with the first pattern hitting
Figure 2 shows two HTFs overhead resistance at the ultimate high.
identified in the study. The
trend start point was deter- National Semiconductor Corp. (NSM), daily
mined by finding a 20-per- Ultimate high 84
cent reversal of the existing 76
68
trend, measured from a 62
prior low to the most recent 56
close. The ultimate high was 50
identified by finding a sub- 44
sequent 20-percent trend HTF 2 40
change, measured from a 36
32
prior high to the recent Ultimate high
28
close.
HTF 1 was preceded by a Trend start 24
22
156-percent rally that lasted 20
40 days. The flag retraced 38 18
HTF 1
percent of the rally and last- 16
ed 15 days. After the break- 14
out, price climbed 54 percent 12
to the ultimate high, which
occurred at a resistance area 10
established by price peaks as Trend start
8
far back as mid-1995 (not
shown).
Price rose 121 percent 6
1999 May June July Aug. Sept. Oct. Nov. Dec. 2000 Feb. Mar.
leading up to HTF 2, taking
Source: Proprietary software (Thomas Bulkowski)
51 days to make the climb.

ACTIVE TRADER • December 2004 • www.activetradermag.com 29


FIGURE 3 TWO MORE HTFS
The first HTF launches from a flat base and price soars to the ultimate high. The second
trade is more typical with the rise to the ultimate high about half the distance, on a per-
centage basis, from the trend start to the flag.

Vertex Pharmaceuticals (VRTX), daily The flag retraced 45 percent


93
85 of the rally and was 29 days
77 long. After the breakout,
Ultimate high 71 price rallied 92 percent. This
65 pattern did not have over-
59
head resistance to overcome
53
49 on its way to the ultimate
Ultimate high
45 high. That may explain why
HTF 2 41 price nearly doubled before
37 trending downward.
33 Notice the irregular
29 shapes of these two HTFs.
25 The first looks like a small
23 broadening wedge and the
21 second looks like a regular
HTF 1 19 falling wedge. Volume
Trend start
17 trends downward in both.
15 Figure 3 shows two more
13 examples. Starting from a
Trend start flat base in late 1999, price
11 climbs 116 percent leading to
HTF 1 and soars 129 percent
9
1999 Dec. 2000 Feb. Mar. Apr. May June July Aug. Sept. afterward.
Source: Proprietary software (Thomas Bulkowski) The stock did not perform
as well after HTF 2 because
the market changed from
FIGURE 4 HTF FAILURE bull to bear between the two
This HTF fails to travel far due to overhead resistance and a change in company patterns (the bear market
fundamentals. The stock tumbles on an earnings warning. started in March 2000, near
HTF 1’s ultimate high). The
Noven Pharmaceuticals (NOVN), daily second HTF pattern was
44
42 preceded by a 136-percent
Resistance Ultimate high
40 price rise and followed by a
38 61-percent rally after the
36 breakout.
34
HTF 32 The measure rule
30 The second pattern in Figure
Dead-cat bounce
28 3 is typical of the rise you
26 can expect after an HTF in a
bull market. For all 252 pat-
24
23 terns in the study, the climb
22 leading to the pattern aver-
21
20
aged 124 percent, but the
19 post-breakout gain was just
18 69 percent.
17 To determine an approxi-
Trend start 16 mate target, compute the
15 percentage change from the
14 low of the trend start point
to the high at the top of the
13
flag. After the breakout, the
12 move from the flag’s lowest
2000 Feb. Mar. Apr. May June July Aug. Sept. Oct. Nov.
low should measure
Source: Proprietary software (Thomas Bulkowski) approximately half this

30 www.activetradermag.com • December 2004 • ACTIVE TRADER


An HTF triggers a buy signal after a stock high, raise the stop to just below the
prior minor low, provided it is not too
has made a significant up move and, thus, far away; otherwise use 1.5 times the
daily volatility, which is 1.5 times the
will appear overbought to many traders. average intraday trading range over the
prior month. Keep raising the stop as
price climbs. Use the measure rule to
amount. This “measure rule” works 90 close above the highest high in the pat- find a target price (half the price rise
percent of the time in a bull market. tern as the buy signal. This is important: leading to the HTF, projected upward
If you buy before the breakout, price from the flag low). Most of the gains (35
Pattern failures might drop instead. Only 10 percent of percent, on average) will come in the
Figure 4 illustrates two types of pattern the 252 patterns in the study failed to first week, so enter as soon as you get the
failures. The first is a rise blocked by climb at least 20 percent, and none failed signal.
overhead resistance. Underlying sup- to climb less than 5 percent. Those are Also, the time from the trend start to
port or overhead resistance (look for a very low failure rates. the flag start will be slightly less (by six
solid mass of horizontal price movement For protection, use progressive stops. days on average) than the time from the
or peaks and valleys stopping near the For example, once price makes a new flag’s end to the ultimate high. 
same price area) spells death to most
chart-pattern breakout trades. In this
example, price climbed 31 percent (the
pre-pattern rally was 128 percent) after
the HTF breakout. That is a significant
rally by most standards, but it fails to
Additional reading
come close to the 64-percent gain using Books:
the measure rule. How to Make Money in Stocks (McGraw-Hill, 1988) by William O’Neil
The second failure comes from the fun-
damentals. The company issued an earn- Books by Thomas Bulkowski:
ings warning for the quarter and said full- Encyclopedia of Chart Patterns (John Wiley & Sons, 2000)
year earnings would suffer as well. The
stock tumbled 43 percent in one session. Trading Classic Chart Patterns (John Wiley & Sons, 2002)
Price bounced up during the next month Active Trader articles:
before “rounding over” and making a
lower low in a classic “dead-cat bounce” “Trading ‘busted’ patterns,” November 2004, p. 42
(DCB) pattern. Three months later, the “Half-staff patterns: Profiting from flags and pennants,” September 2004, p. 48
stock dropped another 32 percent on a
warning about flat annual revenues. “Three falling peaks: Bearish trend change pattern,” August 2004, p. 32
Problems cannot always be fixed in “Chart patterns: Does size matter,” June 2004, p. 44
one quarter. Avoid a stock showing a
“Trading disaster: the dead-cat bounce,” May 2004, p. 44
DCB for at least six months — preferably
a year. That will give the company time “Broadening patterns: Clues to breakout direction,” April 2004, p. 36
to get its act together.
“Technicals meet fundamentals in the earnings flag,” February 2004, p. 30
Trading the pattern “A different breed of scallop,” January 2004, p. 32
An HTF triggers a buy signal after a “The three rising valleys pattern,” December 2003, p. 28
stock has made a significant up move
and, thus, will appear “overbought” to “Pipe bottom reversals,” November 2003, p. 28
many traders. This is a momentum play, “Grabbing the bull by the horns,” September 2003, p. 46
buying high and selling higher. Trading
an HTF is like standing on the edge of the “Head-and-shoulders bottoms: More than meets the eye,” August 2003, p. 32
cliff and jumping off, hoping the water at “Tom Bulkowski’s scientific approach,” September 2002, p. 32
the bottom is deep enough. You need a
good dose of courage to take the plunge. You can purchase past articles at www.activetradermag.com/purchase_arti-
To trade the pattern, wait for price to cles.htm and download them to your computer.
either close above the flag trendline or, if
the HTF has an irregular shape, use a

ACTIVE TRADER • December 2004 • www.activetradermag.com 31


TRADING Strategies

Mastering
TWO-MINUTE breakouts
How can you find consistent trade opportunities?
One way is to trade breakouts through yesterday’s high and low —
but only after the stock has shown its true colors.

trade is to buy upside or downside stock, you can ride the coattails of the
breakouts of the previous day’s high or large money on the way up. To make
low, respectively, avoiding trades in the sure, however, don’t enter the trade until
middle of the day’s range. We’ll show the stock also has cleared the required
how to apply this technique using two- noise level.
minute charts. We also use the time and sales win-
dow to confirm that any large block
The tools trades are going our way and that most
For this approach, use a two-minute can- trades are executed at the ask price for
dlestick chart encompassing a two-day long trades, or the bid price for short
time horizon (today and yesterday). For sales. It also is good if a directional chart
a long trade, buy the stock once it has pattern — i.e., one that implies a move
cleared the whole number closest to the either up or down — confirms the break-
BY KEN CALHOUN previous day’s high. out. A simple example is successive clos-
For example, assume a stock made a es at the high (or low) of the price bars
high of 47.9 yesterday. In this case, you leading up to, or coinciding with, the

I
would enter a buy order when the stock breakout. Also, many traders use specif-
t is often a struggle to find the hits 48.5 (having cleared 48, the nearest ic candlestick chart patterns to indicate
most appropriate indicator for a whole number) and when time and likely price direction.
given trading situation. A tool sales shows that most trades are being
that works in one environment executed at the ask price, which would The rules
may not be appropriate in another. suggest strong demand for the stock. The best time to use this method is the
Momentum oscillators or the Nasdaq Reverse the logic for short trades. profitable and volatile 9:40 a.m. to 11
and S&P 500 futures may provide early The reason for placing the entry a cer- a.m. (EST) time period. Trades typically
signals of shifts in the stock market, but tain amount above the previous day’s last several to 20 minutes. Here are step-
these tools also are often unreliable. high — in this case, 48.5 — is to make by-step guidelines for applying this
Moving average crossovers provide sure the trade safely “clears the hurdle”
trend confirmation but generally lag of the previous day’s trading range,
price action, and you cannot count on accounting for any market noise that
Strategy snapshot
sustained trends in consolidating mar- may be present. We don’t want to buy a Strategy: “Two-day” breakout
kets. Further, market makers frequently double top, we want to buy a breakout Market: Stocks
disguise their intentions via Electronic above the previous day’s high. Entering
Communications Networks (ECNs) or 0.3 to 0.5 above the whole number helps Entry: Go long (short) on move
Level II head fakes, which render the avoid false breakouts. .3 to .5-points above (below) whole
Level II screen more or less useless. So, This approach works because many number closest to previous day’s high
what’s a trader to do? professional traders and institutional (low).
Watch price action. Trading breakouts buyers buy such breakouts. In addition, Exit: Exit with trailing stop or on close.
and breakdowns of chart patterns is a some institutional buy programs also Risk control: Stop-loss of no more than
reliable and simple trading technique factor in the open, high, low and closing 0.4 points. Trail stop at this interval
that can help you limit risk. A relatively prices. When such programs trigger buy if market moves in direction of trade.
consistent short-term, pattern-based signals and money starts flowing into a

32 www.activetradermag.com • September 2001 • ACTIVE TRADER


technique. back into the trading range near the
1. Define the day’s breakout and whole number and the entry price level
breakdown entry levels before each mar- is violated.
ket open. Set up one of your trading 4. Trail the stop to protect profits.
screens to plot a single, large two-minute 5. Because the market often reverses
candlestick chart covering two days around 10 a.m. each day, it is useful to
(today and the previous trading day) of tighten the stop during this time to three
trading activity, as shown in Figure 1. or four “spreads” (the colored bands of
Make sure you start charting by 8:30 bid and ask levels on the Level II screen)
a.m. so you can spot any pre-market top behind the current inside bid. With deci-
or bottom formations, price gaps and mal trading, this allows active traders to
trends. Identify the previous day’s high keep even tighter stops than was previ-
and low. ously possible. Glossary
2. Enter 0.3 to 0.5 points above the pre-
vious day’s high (for long trades) or low Trade examples Time and sales:
(for short trades). Figure 1 shows that on April 27, Ebay The real-time, official record of
3. All intraday trades should have a (EBAY) made a high of 48 and a low of executed trades (as opposed to
maximum stop-loss of 0.4 points. 45.5. Based on the guideline to place the bids and offers) throughout the
Combined with entering 0.3 to 0.5 points entry points 0.3 to 0.5 points above or day. Most trading platforms
above the previous day’s high, this pro- below the previous day’s high and low include a time and sales window
vides an excellent risk management tool. prices, on April 30 we set long entry at to monitor this activity.
In effect, we will exit if the reason for the 48.5 (0.5 points above the previous day’s
trade is negated, i.e., the stock moves high of 48, which was a whole number). Noise:
Random, meaningless price fluc-
tuations that can knock traders
FIGURE 1 BUY SIGNAL
out of the market.
A buy signal occurs in EBAY when the stock moves .5 points above the whole
number nearest to yesterday’s high. Buy programs (program trading):
Computer-based trading
EBay Corp. (EBAY), two-minute approach whereby institutions or
52.00 large trading operations execute
Buy signal is generated
51.50 large volume in related markets
when price exceeds to take advantage of discrepan-
previous day’s high +.5 points. 51.00 cies between them (i.e., buying
S&P stocks and selling S&P
50.50 futures). See “Program trading
0
and fair value,” Active Trader,
50.00
Jan./Feb. 2001, p. 28, for more
49.50 information.
Previous day’s
49.00 Uptick rule:
high: 48.00
Securities and Exchange
48.50
Commission rule that requires
48.00 short sales to be executed when
the last recorded price in a stock
47.50 is higher than (or equal to,
depending on the circumstances)
47.00 the immediately preceding price.
46.50
(The rule varies slightly for NYSE
and Nasdaq stocks, although the
46.00 principle is the same.) See “A
walk on the short side,” Active
45.50 Trader, July 2000, p. 32, for
Previous day Current day more information.
45.00
(compressed)
9:30 10:00 10:30 11:00 11:30 12:00
4/27/01 4/30/01
Source: Data Broadcasting Corp.

ACTIVE TRADER • September 2001 • www.activetradermag.com 33


We trailed a stop no more than 0.4 wrong side when the market opens. entering multiple trades.
points behind the current price level. In Therefore, we adjusted the entry to 41.4 Because the stock already has traded
this trade, the trailing stop was triggered to clear the gap with as small a distance at or close to this price in the pre-market,
at 49.375, yielding a net profit of 0.875 as possible. This is not an exact science. it also is important that the time and
points in less than 20 minutes. Sometimes you will jump into a trade sales window confirms large block
Figure 2 (left) is an example on the too soon despite this step; other times trades are going our way and that most
short side of the market. Adobe (ADBE) this precaution will save you from tak- trades are being executed at the bid price
traded between 42.4 and 45.4 on May 2. ing an unnecessary loss. (indicating selling pressure).
The next day (May 3) we therefore Because of the uptick rule, it may take After the entry at 41.4, we trailed a
looked to go short if the market fell to several attempts to execute a short trade. stop a few spreads behind the open
41.6, 0.4 points below the whole number Don’t be afraid to hit the short button on trade, without exceeding the 0.3-point
(42) closest to the previous day’s low. your trading platform software several stop we’ve set for this trade. Note that
However, ADBE gapped down to 41.6 times (assuming you are using a direct- the stop is slightly tighter in this trade
in pre-market trading. When this hap- access broker) so you can get in on an than in the first example. Because of the
pens, it’s a good idea to move the initial uptick. Check your trade confirmation support-resistance level created by the
entry point farther away from the price window to make sure you are executing pre-market gap to 41.6, this trade will be
action to avoid being caught on the a single short trade, and not mistakenly invalidated as soon as the market trades
above this level, which will happen at
FIGURE 2 SELL SIGNAL 41.7 — 0.3 points away from the entry
price. Most trades entered before 10 a.m.
ADBE had already reached the pre-determined entry price in pre-market should not last any longer than five to
trading. The stock kicked off the official trading session with a two-minute eight minutes. Trades entered after 10
rally. Had we sold immediately on the open without adjusting the entry price a.m. can last a little longer, but never
to take this price action into account, we would have been stopped out with more than 20 minutes. This trade was
a loss. covered at 40.75 for a .65-point profit.
Adobe Systems Inc. (ADBE), two-minute
45.60 Bottom line
45.40 Successful trading is much more difficult
45.20 than it first appears. It requires a long
45.00 process of market watching and practic-
44.80
ing chart pattern recognition. In time,
you can learn to avoid low-potential sit-
44.60
uations and focus on entries based on
44.40
specific chart pattern breakouts and
44.20
breakdowns.
44.00
Planning ahead to trade breakouts
43.80
should be done daily using the previous
43.60 day’s high and low to set trade alerts.
43.40 Trading with the trend on breakouts using
Short signal is
43.20 these criteria will help traders avoid over-
generated at
43.00 trading and selectively trade the strongest
41.40.
42.80 and most powerful chart patterns.
42.60 The only exceptions to trading break-
42.40 outs of the previous day’s trading range
42.20 are those rare occasions when a stock
42.00 makes a rapid multi-point drop from the
41.80 previous day’s high and bounces off the
41.60 previous day’s low. But this is a trade for
Previous day’s 41.40 experienced traders only, and you
low: 42.20 41.20 should not expect to capture more than
41.00 50 percent of the retracements following
40.80
the bounce.
In fact, buying bottoms and shorting
40.60
tops is largely a failing method, despite the
40.40
amazing predisposition of most new
40.20
traders to attempt these types of trades.
40.00
Previous day Current day Your trades should be at least 80 percent
39.80
(compressed) breakouts and no more than 20 percent
bottom bounces, not the other way around.
9:30 10:00 10:30 11:00 11:30 12:00 It’s a good idea to tape that to your
5/2/01 5/3/01 monitor, along with the words, “Tight
Source: Data Broadcasting Corp. stops — no exceptions!” 

34 www.activetradermag.com • September 2001 • ACTIVE TRADER


TRADING Strategies

Swing trading 10-day


CHANNEL BREAKOUTS
To trade breakouts successfully, you have to line up as many market factors
as possible. Incorporating volume and momentum into your trading plan can put
you on the inside track to breakout trades that won’t break apart.

BY KEN CALHOUN

T rying to outguess the mar-


ket by picking bottoms and
tops is usually unsuccess-
ful, and more often than not
results in a large numbers of whipsaw
trades. By contrast, professional traders
and institutions favor breakout trading.
the Securities and Exchange Commission
(SEC) raised the minimum margin
requirement for pattern day traders
(PDTs) to $25,000 on Sept. 28, 2001 (see
“New rules for the intraday trader,”
Active Trader, October 2001). Traders with
less than the $25,000 can make no more
large an upfront investment in capital,
software or equipment. Because it does
not require a trader to watch the market
all day, swing trading can be done on a
part-time basis using online discount
brokers. Professional day trading
requires a full-time commitment and a
Combining 10-day support and resist- than four intraday trades in a five-day fast direct-access broker.
ance lines with confirming signals such period; those who exceed this limit must This makes swing trading a viable
as volume breakouts and reversals is a meet the new day trading margin alternative for active traders who are
practical approach to identifying swing requirements or face potential position unwilling to meet the new margin
trade opportunities. These 10-day “chan- liquidation or account closure. requirements and/or uncomfortable
nels“ provide clear criteria for entering Day trading, in which trades often are with the technology and capital demands
breakout trades once these price levels entered and exited in a matter of seconds, of day trading. Swing trading is also an
are triggered. can be highly stressful and requires a sig- effective way to learn many of the “clas-
nificant initial investment — not just in sic” technical indicators and limit risk
Why swing trading? trading funds, but in computer hardware with small-share or paper trades.
Swing trading is a shorter-term trading and software, and training in direct- The following strategy uses simple
style in which positions are held any- access trading methods as well. volume and sector-strength filters to
where from one to 10 days. Swing trading Swing trading, by contrast, is general- determine when to trade breakouts of
has been increasingly popular ever since ly less stressful and does not require as 10-day price channels.

35 www.activetradermag.com • March 2002 • ACTIVE TRADER


If a stock gaps
The tools The rules
For this approach, use a 15-minute chart The best types of stocks to trade with this open more than
encompassing the most recent 10 days approach are Nasdaq or NYSE stocks
(i.e., today and the previous nine trading priced between $5 and $60, with average
sessions). The following rules are given daily volume of at least 800,000 shares, 10 to 15 percent
in terms of upside breakouts and long and average intraday trading ranges of 1
trades; reverse the rules for short trades.
However, this strategy is better for long
to 4 points. Here are the rules: from its previous
swing entries. 1. Define the 10-day high and low for
For a breakout swing trade, buy when the stock using a 15-minute candlestick close, it will often
a stock breaks out at least 50 cents over or bar chart. Be sure to include volume
the whole number above the 10-day
high, accompanied by volume that is
bars on the chart.
reverse and fill
higher than the previous day’s volume at 2. Enter 50 to 60 cents above the near-
the same time. For example, assume that est whole number above the highest high
of the past 10 days, including today.
the gap, in which
during the previous nine trading ses-
sions plus today, the highest price a stock
traded at was 37.8, which was set on the 3. Look for volume breakouts on the case it’s necessary
previous trading day. The trigger for a 10-day chart. Compare volume bars on
10-day breakout long trade would be
38.5, as long as the volume in the current
the current trading day to previous trad-
ing days. The best entries are those for
to take your profit
session is higher than it was at the same which volume is higher than in the pre-
time in the previous session. If the stock vious session. before the market
opened today at 37.6 and traded up to
38.5, this would trigger a long trade. 4. Confirm entries using market indi-
The only exception is when the entry cators such as the Arms Index (TRIN) as does.
price would contain a “9” — e.g., 19.5, well as the time of day.
29.5, 39.5, and so on; in such cases, wait The TRIN measures the net buying The TRIN, versions of which are avail-
until the stock clears the nearest multiple pressure vs. selling pressure in the mar- able for both NYSE and Nasdaq stocks,
of 10, which would result in long trade ket at a given point in the trading day. can help determine whether a trade is
triggers at 20.5, 30.5, 40.5, etc. The ration- The formula is: advisable by highlighting whether
ale is prices with a “9” tend to look expen- momentum is bullish or bearish at a
sive and often meet resistance, choppy {number of advancing issues/number given time. A TRIN reading of 1 means
price action, or both near multiples of 10. of declining issues}/{volume of advanc- buying/selling volume and the number
ing issues/volume of declining issues} of advancers/decliners are equally
matched; a TRIN reading above 1 is
bearish; a TRIN under 1 is bullish. See
Strategy snapshot Indicator Insight, Active Trader,
Strategy: 10-day channel breakout December 2000, for more information on
Markets: Nasdaq or NYSE stocks trading between $5-$60, this indicator.
with average daily volume of at least 800,000 shares It’s also helpful to enter at times of the
and average daily range of 1 to 4 points. day when the market is the strongest
and most volatile. It’s usually best to
Entry (for longs; Go long 50 to 60 cents above the nearest whole number enter 10-day channel long trades in the
reverse for shorts): above the highest high of the past 10 days. early morning, from 9:45 until 11 a.m.
Confirmation: The best entries are those in which volume is higher than EST, or during late-afternoon rallies —
in the previous session and/or when the stock is in between 2:30 and 3:30 p.m., for example.
a strongly trading sector. Avoid highs that are reached on Certain cautionary indicators (“red
lower-than-average volume or those reached by a stock in flags”) can be used to eliminate poor
a weak sector on the entry day. trades. For long entries, avoid highs
Exit/risk control: The widest initial stop should be the closer of 1.5 points reached on lower-than-average volume
or the previous day’s low. The most conservative initial or those reached by a stock in a weak sec-
stop-loss is the previous day’s high. Once in a profitable tor that day. Compare sector indices such
trade, trail the stop .5 points below the current trading as the SOX, NBI, GHA and GSO to deter-
range. mine which are strongest, and give pref-
erence to entries in the strongest sectors

ACTIVE TRADER • March 2002 • www.activetradermag.com 36


FIGURE 1 POISED TO BREAK OUT
your risk tolerance.
NVDA traded in a channel from 48 to 55 for 10 days, forming an ascending A good initial stop-loss is the previous
triangle toward the end of the period as it challenged the resistance level day’s low or 1.5 points, whichever is
another time. Entry would occur at 55.50, 50 cents above the highest high smaller. For example, let’s say the 10-day
of the past 10 days. channel range for EBAY is bounded by a
low (support level) of 61.8 and a high
Nvidia (NVDA), 15-minute 55.50 (resistance level) of 66.8. If the previous
Resistance 55.00
54.50 day’s range was 65.5 to 66.8, we would
54.00 enter EBAY on a breakout above 67.5.
53.50 The initial maximum stop-loss for this
53.00
52.50 trade would be at 66, 1.5 points below
52.00 entry (roughly 2 percent).
51.50
51.00
If you trade on a shorter time frame,
Ascending 50.50 say three- or five-minute charts, you
triangle 50.00 might consider setting a tighter stop at
49.50
49.00 the previous day’s high.
Support 48.50
48.00 6. Trail a stop to protect open profits at
Volume 2 percent (generally from .5 to 1.5 points)
6 million below the current level of the open trade,
4 million or use a time stop of no longer than 10
2 million
days (i.e., exit all remaining open posi-
12:00 12:00 12:00 12:00 12:00 12:00 12:00 12:00 12:00 12:00
tions after 10 days). Whenever one of the
11/16/01 11/19/01 11/20/01 11/21/01 11/23/01 11/26/01 11/27/01 11/28/01 11/29/01 11/30/01 exit signals appears, the position should
Source: eSignal be closed with a profit. Re-enter on sub-
sequent breakouts after retracements
have occurred.
FIGURE 2 AFTER THE BREAKOUT
How to handle gaps
After the stock fulfills the entry requirements and breaks out above
Managing gap opens on swing trades is
resistance, a trailing stop is used to lock in profits.
always a challenge. When a stock gaps
Intuit Inc. (INTU), 15-minute open significantly above the previous
44.00 day’s high (in your favor for a long
Entry 43.50 swing trade), trail a stop no more than 50
43.00 cents below the current pre-market trad-
42.50 ing range to lock in your profit.
Resistance
42.00 However, if a stock gaps open more
41.50 than 10 to 15 percent from its previous
41.00 close, it will frequently reverse and fill
40.50 the gap, in which case it’s necessary to
40.00 take your profit before the market does.
39.50 This is especially true if the stock gaps
Support 39.00 up above the previous day’s high.
Conversely, when a stock gaps down
Volume
600,000 significantly against you, it’s often best
400,000 to wait until 15 to 20 minutes after the
200,000 open to exit the position, because down
gaps frequently attract buyers who can
12:00 12:00 12:00 12:00 12:00 12:00 12:00 12:00 12:00 12:00 bring the price back up. Again, use a
11/16/01 11/19/01 11/20/01 11/21/01 11/23/01 11/26/01 11/27/01 11/28/01 11/29/01 11/30/01
stop-loss of no more than 50 cents below
Source: eSignal the current pre-market trading range.
It is sometimes helpful to wait until
— e.g., those up 1.5 to 3 percent or so on Mixed, choppy days are poor days for approximately 9:45 a.m. EST to see
the day at the time of the trade entry. swing trade entries. where the stock trades before exiting a
Also check to see if sectors are conver- position. It is frustrating to panic out of a
gent (all green or red — i.e., moving up 5. Stop-loss values are determined by gap-down swing trade only to see the
or down) or divergent (mixed). It’s best the previous day’s high and low. Either stock turn around and fill the gap in the
to enter swing trades on days where all of these price points can provide you first few minutes of the trading day.
sectors are convergent and the broad with an initial stop-loss value, depend- Calmly give it a few minutes to establish
market has strength in a single direction. ing on the intraday market trend and a trend and see if it consolidates and

37 www.activetradermag.com • March 2002 • ACTIVE TRADER


it is sound risk management to “extend
FIGURE 3 OVERFLOWING CUP
yourself” only on the strongest of pat-
Like channels and ascending triangles, cup patterns also provide well-defined terns. This keeps the average entry price
resistance levels for breakout trades. The stock broke out above a second cup toward the low end of the total position.
pattern on Dec. 7, and was stopped out with a 3.6 point profit on Dec. 11. Cup-pattern breakouts, like ascend-
ing-triangle and consolidation breakouts,
40.00
Invision Technologies (INVN), 15-minute are much stronger than cases where a
38.00
stock simply trades to a new high with-
36.00 out penetrating any kind of resistance
34.00 level. The test of sellers that occurs at a
Breakout
entry
32.00 resistance level prior to a long cup-pat-
Resistance 30.00 tern breakout validates the entry and
Exit
28.00 provides a support level after the trade.
26.00 Cup patterns, which are extended,
24.00 saucer-shaped retracements, appear fair-
22.00
ly frequently. The key to trading them is
Initial cup Second cup to apply volume and other filters to
20.00
avoid false breakouts that turn out to be
18.00
double tops (i.e., when price falls back
16.00 from the right side of the cup instead of
Volume 1 million breaking out above the resistance level
of the cup).
500,000 Figure 3 shows a cup pattern that
started on Dec. 3, pulled back from the
resistance level of 28.59 on Dec. 5 (form-
12:00 12:00 12:00 12:00 12:00 12:00 12:00 12:00 12:00 12:00
ing a short-lived double top), formed
11/29/01 11/30/01 12/3/01 12/4/01 12/5/01 12/6/01 12/7/01 12/10/01 12/11/01 12/12/01
another cup and finally broke out above
Source: eSignal the resistance on the afternoon of Dec. 7.
A long trade was triggered at 30.5
reverses this initial gap move. Figure 2 provides an example of how (because the entry price would have
to manage a profitable long swing trade. been 29.5, and in such cases entry is
Trade examples On the afternoon of Nov. 29, 2001, the made above the nearest multiple of 10).
Figure 1 shows Nvidia (NVDA) trading stock cleared the 10-day high and a long The stock gapped open higher on Dec.
in a 10-day channel (between 48 to 55) trade was entered at 42.50. The initial 10 and the trade was exited on Dec. 11
from Nov. 16, 2001, to Nov. 30, stop loss was set at 41 (42.5-1.5). (when the trailing stop was hit) at 34.1
2001. Based on the guideline to enter 50 On Nov. 30, the stock continued to for a 3.6-point profit.
cents over the nearest higher whole num- rally throughout the session to a high of
ber above the highest high of the 10 days, 44, at which point it consolidated. Using Bottom line
a long entry would be triggered at 55.5. a trailing stop approach raises the stop Swing trading provides traders with
(If, however, the stock gaps up to, say, to 43.5 (44-.5), which was not triggered. opportunities to manage multiple posi-
55.8 in pre-market trading, the entry The “time stop” is 10 days from Nov. 29. tions and entries at a more leisurely trad-
would be reset over the next number up, In this case the stock is up more than a ing pace than is possible in the hectic
at 56.5 or higher). point on an overnight hold. We continue world of day trading. However, every
The initial stop would be placed at 54, to trail a stop .5 behind the current trad- trader should research and experiment
1.5 points below the entry, because 1.5 ing range until the stop is taken out. with different trading styles to help
points is a tighter stop than the previous determine his or her preference and level
day’s low (see trade rule No. 5). The Dynamic position sizing of comfort.
alternate, more conservative stop-loss using cup breakouts Using 10-day trading channels to iden-
level is the previous day’s high — in this Dynamic position sizing is the process of tify entries on volume breakouts can help
case 54.60. adding to an initial position once a stock you better define support and resistance
Notice this stock forms an ascending has broken out and is continuing to levels and provide techniques you can
triangle pattern following an initial attract buyers. For example, a trader integrate with other technical indicators
downturn earlier in the 10-day channel, may buy 200 shares initially and add to develop a swing-trading plan.
and is poised to break out to new highs another 100 shares on a subsequent Using a comprehensive, measured and
if it clears the 55 resistance area. If the breakout as the stock continues to climb. specific strategy to trade breakouts con-
volume when the trade is entered (on The key to using dynamic position tinues to produce entries that are more
Dec. 1, not shown) is higher than it was sizing is to add no more than half the consistent than intra-range or “bounce”
at the same time on the previous day, number of the initial trade size on subse- trade approaches. Using volume and
this would provide additional confirma- quent 10-day high cup-pattern break- price action filters will help you avoid
tion for a long trade. outs. When adding to an initial position, false breakouts in choppy markets. 

ACTIVE TRADER • March 2002 • www.activetradermag.com 38


EQUITY CURVE
3,000,000

2,500,000
Volatility breakout system

Account balance ($)


2,000,000
System logic: The volatility breakout system is a classic
trading strategy based on identifying situations when a 1,500,000
market is about to burst out of a congestion area and poten-
tially establish a new long-term trend. It also can signal a 1,000,000
trade if the market is already trending in one direction but
quickly reverses to establish a new trend in the opposite 500,000
direction.
This system uses Bollinger Bands (complemented by
0
moving averages), which are lines typically plotted two 12/7/92 12/7/93 12/7/94 12/7/95 12/7/96 12/7/97 12/7/98 12/7/99 12/7/00 12/7/01
standard deviations above and below a moving average.
Bollinger Bands expand during high-volatility periods and contract
during low-volatility periods. Test data: Daily prices for 14 Dow Jones Industrial Average stocks
When volatility is high, the system is designed to stay out of the (AXP, C, CAT, DIS, GM, HWP, IBM, INTC, JPM, KO, MO, MRK,
market to avoid taking any unnecessary risk, but if an entry is trig- MSFT, T), with $10 deducted per trade for slippage and commis-
gered anyway, the system will work to keep you in the trade to sions.
avoid being stopped out prematurely with a loss. A long entry is
triggered when price moves above its 60-day moving average and Starting equity: $1 million (nominal).
breaks the upper Bollinger Band. To exit, price must move below its
30-day moving average and break a lower Bollinger Band that is set Buy-and-hold stats:
one standard deviation away (instead of the usual two). Because the Total Maximum Longest
next entry cannot occur until price moves back above the lower Index return drawdown flat period
band, above its moving average and above the upper band, there DJIA 175% 31.5% (current) 29 months (current)
will be times when the system is out of the market completely. S&P 500 120% 40% (current) 26 months (current)
Nasdaq 182% 80.5% (current) 26 months (current)
Markets: This system will be tested on stocks and also on futures
(p. 70). Test results: The system was originally tested on all 30 stocks in
the DJIA; the 14 in this test were selected because they were the
Rules: ones that showed a profit. Singling out these stocks, though,
1. Go long tomorrow if price moves above its average price for the
last 60 days and breaks the
upper Bollinger Band. SAMPLE SIGNALS
2. Exit with a profit or loss if
Philip Morris (MO), daily 58.00
price moves below its 30-day
moving average and penetrates LX = Long exit
LX 56.00
a lower Bollinger Band set one
standard deviation away.
LX Buy 54.00
Reverse the rules for short Sell
trades. 52.00

Money management: 50.00


1. Risk 4 percent of available
equity per stock traded.
2. The number of shares to trade Buy 48.00
(ST) is calculated using the fol-
lowing formula: 46.00

ST = AC * PR / R
where 44.00
AC = Available capital
PR = Percent risked
R = Distance between entry 42.00
price and exit price (stop-loss).

Test period: November 1992 to February March April May June July
June 2002. Source: Omega Research ProSuite

39 www.activetradermag.com • October 2002 • ACTIVE TRADER


DRAWDOWN CURVE
12/7/92 12/7/93 12/7/94 12/7/95 12/7/96 12/7/97 12/7/98 12/7/99 12/7/00 12/7/01
0%
In this system, though, the large distance between
-5% the entry and exit prices requires trading rather small
positions. Doing otherwise would run the risk of
-10% using all the capital on only a few positions. This in
turn will result in relatively small dollar gains despite
large price swings. Even though the system trades
-15%
only 14 stocks, close to 75 percent of available capital
is tied up.
-20% Also, the current drawdown has gone on for 42
months. This is likely a reflection of the disappearance
-25% of the stock market’s pre-2000 trending characteristics,
and it is not very likely that a system like this will start
-30% producing a profit anytime soon.
This system is also a bit passive in its trade fre-
failed to create results as good as the ones produced by the test on quency. To make it more aggressive, the lookback period can be
currency futures (see Futures System Lab, p. 44). shortened and/or the standard deviation boundaries
One thing to keep in mind is that portfolio composition is more tightened.
complex than simply eliminating instruments that don’t perform
well. In a dynamic portfolio such as this one, where a group of
stocks interacts within a single trading account, the ability of an ROLLING TIME WINDOW RETURN ANALYSIS
individual stock to turn a profit or loss also depends on the behav- Cumulative 12 24 36 48 60
ior of all the other stocks. months months months months months
As proof of this, three stocks that were profitable when all 30
Most recent: 16.70% -0.91% -5.86% 15.22% 22.04%
stocks were tested showed a loss when the field was pared to 14. If
Average: 10.76% 24.50% 42.81% 64.40% 83.77%
we were to test only the remaining 11 that showed a profit, it’s like-
ly a few more would turn into losers. Best: 58.88% 87.95% 102.78% 162.29% 155.19%
This correlation also makes it extremely difficult to trade a long- Worst: -15.08% -19.33% -16.50% 4.57% 16.53%
term system on the stock market. The individual stocks either St. dev.: 17.29% 22.72% 30.43% 36.41% 42.26%
trend well, almost all at once, or they don’t, which results in a large
amount of whipsaw losing trades and rather severe drawdowns. Annualized 12 24 36 48 60
The way to overcome this is to diversify by trading as many stocks months months months months months
from different sectors and groups as possible. Most recent: 16.70% -0.45% -1.99% 3.61% 4.06%
Average: 10.76% 11.58% 12.61% 13.23% 12.94%
STRATEGY SUMMARY Best: 58.88% 37.10% 26.57% 27.26% 20.61%
Profitability Trade statistics Worst: -15.08% -10.19% -5.83% 1.12% 3.11%
End. equity ($): 2,491,172 No. trades: 456 St. dev: 17.29% 10.78% 9.26% 8.07% 7.30%
Total return (%): 149 Avg. trade ($): 3,270
Avg. annual ret. (%): 10.00 Avg. DIT: 35.0 LEGEND: Cumulative returns — Most recent: most recent return from start to
Profit factor: 1.34 Avg. win/loss ($): 31,090 (13,546) end of the respective periods • Average: the average of all cumulative returns
from start to end of the respective periods • Best: the best of all cumulative returns
Avg. tied cap (%): 73 Lrg. win/loss ($): 391,627(109,437) from start to end of the respective periods • Worst: the worst of all cumulative
Win. months (%): 53 Win. trades (%): 38.8 returns from start to end of the respective periods • St. dev: the standard devia-
tion of all cumulative returns from start to end of the respective periods
Drawdown TIM (%): 100 57.7
Annualized returns — The ending equity as a result of the cumulative returns,
Max. DD (%): 25.8 Tr./Mark./Year: 4.3
raised by 1/n, where n is the respective period in number of years
Longest flat (m): 41.5 Tr./Month: 4.0
Send Active Trader your systems
LEGEND: End. equity ($) — equity at the end of test period • Total return If you have a trading system or idea you’d like tested, send it to
(%) — total percentage return over test period • Avg. annual ret. (%) —
us at the Trading System Lab. We’ll test it on a portfolio of
average continuously compounded annual return • Profit factor — gross
stocks or futures (for now, maximum 60 markets, using the last
profit/gross loss • Avg. tied cap (%) — average percent of total available cap-
ital tied up in open positions • Win. months (%) — percentage profitable 2,500 trading days), using true portfolio analysis/optimization.
months over test period • Max. DD (%) — maximum drop in equity • Most system-testing software only allows you to test one mar-
Longest flat — longest period, in months, spent between two equity highs • ket at a time. Our system-testing technique lets all markets
No. trades — number of trades • Avg. trade ($) — amount won or lost by share the same account and is based on the interaction within
the average trade • Avg. DIT— average days in trade • Avg. win/loss ($) the portfolio as a whole.
— average wining and losing trade, respectively • Lrg. win/loss ($) — Start by e-mailing system logic (in TradeStation’s
largest wining and losing trade, respectively • Win. trades (%) — percent EasyLanguage or in an Excel spreadsheet) and a short description
winning trades • TIM (%) — amount of time there is at least one open posi- to editorial@activetradermag.com, and we’ll get back to you.
tion for entire portfolio, and each market, respectively • Tr./Mark./Year — Note: Each system must have a clearly defined stop-loss level
trades per market per year • Tr./Month — trades per month for all markets and a suggested optimal amount to risk per trade.

Disclaimer: The Trading System Lab is intended for educational purposes only to provide a perspective on different market concepts. It is not meant to recommend or
promote any trading system or approach. Traders are advised to do their own research and testing to determine the validity of a trading idea. Past performance does not
guarantee future results; historical testing may not reflect a system’s behavior in real-time trading.

ACTIVE TRADER • October 2002 • www.activetradermag.com 40


FUTURES & OPTIONS
System Lab
EQUITY CURVE
Futures volatility breakout system 5,000,000
4,500,000
System logic: The system uses Bollinger Bands and a moving aver-
age to trade volatility breakouts. The logic and tools for this system 4,000,000
are described in the Trading System Lab on p. 56. 3,500,000

Account balance ($)


Markets: Most trending futures markets, such as currencies, energies 3,000,000
and interest rates. This system was also tested on stocks (see Trading
2,500,000
System Lab).
Rules: 2,000,000
1. Go long tomorrow if price moves above the 60-day moving aver- 1,500,000
age and breaks the upper Bollinger Band. 1,000,000
2. Exit with a profit or loss if price moves below its 30-day moving
average and penetrates a lower Bollinger Band set one standard 500,000
deviation away (instead of the usual two). 0
Reverse the rules for short trades. 11/25/92 11/25/93 11/25/94 11/25/95 11/25/96 11/25/97 11/25/98 11/25/99 11/25/00 11/25/01
Money management
1. Risk the following percentages of available equity per market: 2 The risk for this system was 2 or 4 percent per trade, which result-
percent for Australian dollar, British pound, Canadian dollar, Dollar ed in both a drawdown and flat time that would be deemed unac-
index and Swiss franc, and 4 percent for Japanese yen, D-mark and ceptable by most professional money managers. These numbers
Euro. (The yen and the Euro are traded with twice the risk because should stay under 30 percent and 18 months, respectively.
they are the most liquid currencies.) The major disadvantage of a trend-following system is most mar-
2. The number of contracts to trade (CT) is calculated with the kets that work well with this type of a system are usually correlated.
following formula: In other words, they will either all work well or perform poorly
simultaneously. This is reflected in the erratic look of the system’s
CT = (AC * PR) / (R * PV) equity curve. Because of this, the drawdowns can be both deep and
where long, and it usually requires a couple of very good trades to get the
AC = Available capital system profitable again. Research has shown that a trend-following
PR = Percent risked system will work best when traded on 15 to 20 select markets from
R = Distance between entry price and exit price (stop-loss). various sectors of the economy.
PV = Dollar value of a one-point move. The system ties up an average of 11 percent of capital. This means
there is plenty of room to add markets, and even trade the system
Test period: November 1992 to June 2002 more aggressively, without extending ourselves too much. This is
Test data: Daily futures prices for eight currency futures: Japanese because futures have much smaller margin requirements than
yen, Australian dollar, Canada dollar, British pound, Dollar index, stocks. Theoretically, as many as 40 to 50 different futures contracts
Swiss franc, and D-mark (until Dec. 31, 1999)/Euro (after Dec. 31, could be traded before reaching the same level of margin fewer than
1999). 20 stocks would require.
Starting equity: $1 million (nominal). Finally, the most recent drawdown of approximately 30 percent is
Test results: If there ever was a system built for the currency markets, the only one over the last 10 years of such magnitude. Most of the
this is it. The reason is the smooth, long-term trends currencies some- previous drawdowns bottomed between 10 to 20 percent. Therefore,
times exhibit, probably because they are mostly influenced by the the latest drawdown is quite possibly an anomaly. That said, how-
global, long-term economical and political climate. In contrast, the ever, there are no guarantees in the market, and your worst draw-
stock market is very sensitive to all other markets; while the stock down is always still to come.
market cares a great deal about the currency market, the currency The system also has a relatively low trade frequency. To make it
market doesn’t care all that much about the stock market. more aggressive, the lookback period can be shortened and/or the
standard deviation boundaries tightened.
STRATEGY SUMMARY
ROLLING TIME WINDOW RETURN ANALYSIS
Profitability Trade statistics Cumulative 12 24 36 48 60
End. equity ($): 4,550,290 No. trades: 303 months months months months months
Total return (%): 355 Avg. trade ($): 11,717 Most recent: 34.31% 58.39% 65.73% 114.82% 138.14%
Avg. annual ret. (%): 17.13 Avg. DIT: 35.8 Average: 15.07% 34.15% 55.68% 80.53% 109.16%
Profit factor: 1.32 Avg. win/loss ($): 53,057 (29,536) Best: 74.75% 76.50% 108.66% 137.94% 202.94%
Avg. tied cap (%): 11 Lrg. win/loss ($): 345,600 (131,350) Worst: -28.10% -1.33% 4.12% 28.76% 38.77%
Win. months (%): 51 Win. trades (%): 42.1 St. dev.: 18.17% 16.26% 23.27% 22.74% 37.29%
Drawdown TIM (%): 97 54.1 Annualized 12 24 36 48 60
months months months months months
Max. DD (%): 31.6 Tr./Mark./Year: 4.0 Most recent: 34.31% 25.85% 18.34% 21.06% 18.95%
Longest flat (m): 19.7 Tr./Month: 2.6 Average: 15.07% 15.82% 15.90% 15.91% 15.90%
Best: 74.75% 32.86% 27.78% 24.20% 24.82%
LEGEND: End. equity ($) — equity at the end of test period • Total return (%) — total Worst: -28.10% -0.67% 1.36% 6.52% 6.77%
percentage return over test period • Avg. annual ret. (%) — average continuously com- St. dev: 18.17% 7.82% 7.22% 5.26% 6.54%
pounded annual return • Profit factor — gross profit/gross loss • Avg. tied cap (%) —
average percent of total available capital tied up in open positions • Win. months (%) — LEGEND: Cumulative returns — Most recent: most recent return from start to end of the
percentage profitable months over test period • Max. DD (%) — maximum drop in equi- respective periods • Average: the average of all cumulative returns from start to end of the
ty • Longest flat — longest period, in months, spent between two equity highs • No. respective periods • Best: the best of all cumulative returns from start to end of the respective
trades — number of trades • Avg. trade ($) — amount won or lost by the average trade periods • Worst: the worst of all cumulative returns from start to end of the respective peri-
• Avg. DIT— average days in trade • Avg. win/loss ($) — average wining and losing ods • St. dev: the standard deviation of all cumulative returns from start to end of the respec-
trade, respectively • Lrg. win/loss ($) — largest wining and losing trade, respectively • tive periods
Win. trades (%) — percent winning trades • TIM (%) — amount of time there is at least
one open position for entire portfolio, and each market, respectively • Tr./Mark./Year — Annualized returns — The ending equity as a result of the cumulative returns, raised by 1/n,
trades per market per year • Tr./Month — trades per month for all markets where n is the respective period in number of years

41 www.activetradermag.com • October 2002 • ACTIVE TRADER


TRADING Strategies

Better breakout trading:


THE NOISE CHANNEL SYSTEM

All breakout traders have to deal


with the reality of false moves
BY DENNIS MEYERS, PH.D. and whipsaws. The noise channel
breakout system shows how a filter

P
can improve the performance
rice trends begin with a breakout of a previous
high or previous low. Unfortunately, many break- of intraday breakout trading.
outs are random — mere market noise. False
moves and reversals can repeatedly whipsaw
traders who act immediately on typical breakout signals. In the tests that illustrate this strategy, we’ll use five-minute
As a result, traders sometimes attempt to use filters to bars of IBM from Feb. 21 to April 6. (For an important point on
improve the odds of catching a successful breakout trend. One testing stock trading strategies, see “A note on price data and
example of a simple filter is to wait for consecutive closes dividends,” p. 75). Intraday data has a high noise level, mean-
above or below a breakout level. Another example is waiting ing it contains a great deal of random price movement that
for price to penetrate a breakout level by x percent or points looks significant but turns out to be meaningless. Without
before acting on the signal. some kind of filter, the losses generated by the random price
The following discussion will analyze a variation on the movement (that is, whipsaws) can completely overwhelm a
simple channel breakout system that uses the latter type of fil- trading system. To help eliminate such random movement, we
ter to minimize whipsaws on an intraday basis. The strategy will add a noise filter, designated by the symbol f, to the basic
will be tested on International Business Machines (IBM). The channel breakout system.
discussion is broken into two parts, covering 1) the system There are three system parameters to find:
rules and data selection and 2) testing procedures. This will • nhi, which is the number of bars in the lookback period
give you the necessary tools for performing similar research used to determine the highest high price (hhp).
and tests on other markets. • nlo, which is the number of bars in the lookback period
used to determine the lowest low price (llp).
The noise channel breakout system • f, which is the amount price must exceed the hhp or llp to
The basic system we will use here is a fairly simple and effec- trigger a buy or sell.
tive breakout system that has been in the public domain for
many years: the channel breakout system, which goes long on The rules for the resulting noise channel breakout system
a move above the highest high of the last n bars and goes short (NCBS) are:
on a move below the lowest low of the last n bars. Buy rule: If price crosses above the highest high of the last n
bars (nhi) by an amount greater than or stable — i.e., the TABLE 1 OPTIMUM PARAMETER VALUES FOR TEST DATA
equal to f, buy at market. For example, if profits, wins and
n = 20 and f = 2 (points), you would go drawdowns should
long when price moved 2 points above not change much as Start date End date nhi nlo f
the highest high of the last 20 bars. the parameters move
In addition, when short, and when by a small amount 2/21/01 3/23/01 8 4 1
calculating the highest high price (hhp), away from their opti-
it cannot be higher than the previously mum values. In other 2/28/01 3/30/01 8 4 1
calculated hhp as previous highs are words, the system
dropped out of the lookback window.
Otherwise, a situation can occur where TABLE 2A TEST PERIOD 1
there is a higher hhp without the price
Performance summary for noise channel breakout system: IBM, five-minute
filter f being hit. Therefore, when short
bars from Feb. 21 to March 23. Statistics based upon trading 1,000 shares of
the stock, the hhp can only stay the same
IBM.
or go lower. It cannot go higher.
Sell rule: If price crosses below the Performance summary: All trades
lowest low price of last n bars (nlo) Total net profit ($): 13,890 Open position P/L ($): 0
minus an amount greater than or equal
Gross profit ($): 39,260 Gross loss ($): -25,370
to f, sell at market. In addition, when
long and when calculating the lowest Total no. of trades: 48 Percent profitable (%): 54
low price (llp), it cannot be lower than Number winning trades: 26 Number losing trades: 22
the previous calculated llp as previous Largest winning trade ($): 5,940 Largest losing trade ($): -2,060
lows are dropped out of the lookback
window. Again, to avoid the situation Average winning trade ($): 1,510 Average losing trade ($): -1,153.18
where a lower llp occurs without the Ratio avg. win/avg. loss: 1.309 Avg. trade(win & loss) ($): 289.38
price filter f being hit, when long the Max. consec. winners: 4 Max. consec. losers: 3
stock, the llp can only stay the same or
Avg. no. bars in winners: 39 Avg. no. bars in losers: 21
go higher. It cannot go lower.
Exit rule: Close the position five min- Max intraday drawdown ($): -8,470
utes before the NYSE close (no trades are Profit factor: 1.547 Max. no. contracts held: 1
carried overnight).
TABLE 2B TEST PERIOD 2
Testing the strategy
The “walk-forward testing” approach Performance summary for noise channel breakout system: IBM, five-minute
will be used to test this strategy because bars from Feb. 28 to March 30. Statistics based upon trading 1,000 shares
of the volatile nature of intraday stock of IBM.
prices. Intraday price dynamics are con-
stantly changing because of economic Performance summary: All trades
surprises, events and trader sentiment. Total net profit ($): 10,490 Open position P/L ($): 0
Also, the time of year — such as the sea- Gross profit ($): 35,500 Gross loss ($): -25,010
son, holidays, vacation time, etc. —
Total no. of trades: 47 Percent profitable (%): 47
affects the character of intraday markets.
As a result, tests performed on intraday Number winning trades: 22 Number losing trades: 25
data three months ago may no longer be Largest winning trade ($): 5,940 Largest losing trade ($): -1,840
representative of today’s intraday price Average winning trade ($): 1,613.64 Average losing trade ($): -1,000.40
action. For more information on walk-
Ratio avg. win/avg. loss: 1.613 Avg. trade(win & loss) ($): 223.191
forward testing and how it was used for
this strategy, see “Proper system test- Max. consec. winners: 3 Max. consec. losers: 3
ing,.” Avg. no. bars in winners: 39 Avg. no. bars in losers: 26
The best parameters will be defined as
Max. intraday drawdown ($): -9,660
those values that generate the best net
profits combined with the minimum Profit factor: 1.419 Max. no. contracts held: 1
drawdown and minimum largest losing Source: TradeStation by TradeStation Group Inc.
trades. In addition, the results should be

43 www.activetradermag.com • September 2001 • ACTIVE TRADER


performance using an nhi of 10 bars from March 26 to April 6. The trades in this generated from the optimized parameters
should be similar to that using nine bars time period are the out-of-sample trades from the two test sections of Feb 21 to
or 11 bars. Also, in choosing the “best”
parameters, we considered only those TABLE 3 OUT-OF-SAMPLE RESULTS
results with four or less maximum con- Combined walk-forward out-of-sample performance summary for the noise
secutive losses. channel breakout system: IBM five-minute bars from March 26 to April 6.
Statistics based upon trading 1,000 shares of IBM.
Test results
Table 1 shows the optimum parameter Performance summary: All trades
values for the test window described in
Total net profit ($): 8,390 Open position P/L ($): 0
“Proper system testing.” The nhi was
eight bars, the nlo was four bars and f Gross profit ($): 14,460 Gross loss ($): -6,070
was 1 point. Tables 2a and 2b show test Total no. of trades: 16 Percent profitable (%): 50
results using these parameters. Number winning trades: 8 Number losing trades: 8
Table 3 summarizes the combined
Largest winning trade ($): 4,000 Largest losing trade ($): -1,350
performance of the two out-of-sample
data segments from March 26 to April 6. Average winning trade ($): 1,807.50 Average losing trade ($): -758.75
This performance represents what Ratio avg. win/avg. loss: 2.382 Avg. trade(win & loss) ($): 524.38
would have happened in real time if you
Max. consec. winners: 5 Max. consec. losers: 3
used the system parameters found in the
test section (not including slippage and Avg. no. bars in winners: 54 Avg. no. bars in losers: 37
commissions). By comparison, the same Max. intraday drawdown ($): -4,480
nhi and nlo values tested without any Profit factor: 2.382 Max. no. contracts held: 1
filter resulted in a loss of $1,150.
Table 4 is a trade-by-trade summary Source: TradeStation by TradeStation Group Inc.

TABLE 4 TRADE-BY-TRADE SUMMARY


This trade summary for the out-of-sample test (five-minute bars, March 26 to April 6) of the noise channel breakout
system shows the strategy actually worked better on the short side than the long side.
Entry Entry Buy Entry Exit Exit Exit # bars P&L ($) P&L (%) Max. Time Max. Time
Date time or price date time price in profit drawdown
sell trade ($)
3/26/01 10:20 Sell 93.75 3/26/01 15:55 94.52 67 (770) -0.82% 0 10:20 (1,620) 10:35
3/27/01 10:15 Buy 95.59 3/27/01 15:55 99.59 68 4,000 4.18% 4,300 15:50 0 10:15
3/28/01 9:40 Sell 97.92 3/28/01 15:55 94.50 75 3,420 3.49% 3,420 12:00 (380) 9:40
3/29/01 10:05 Buy 96.05 3/29/01 15:05 94.90 60 (1,150) -1.20% 950 10:30 (1,160) 11:20
3/29/01 15:05 Sell 94.90 3/29/01 15:55 94.88 10 20 0.02% 390 15:15 (500) 15:45
3/30/01 9:40 Buy 96.70 3/30/01 13:05 96.20 41 (500) -0.52% 800 11:55 (1,190) 10:00
3/30/01 13:05 Sell 96.20 3/30/01 15:55 96.25 34 (50) -0.05% 220 13:15 (840) 14:35
4/2/01 9:40 Buy 97.75 4/2/01 10:55 96.40 15 (1,350) -1.38% 350 10:05 (1,350) 10:55
4/2/01 10:55 Sell 96.40 4/2/01 15:55 94.50 60 1,900 1.97% 2,600 15:40 (1,300) 11:40
4/3/01 10:00 Sell 93.00 4/3/01 15:55 90.50 71 2,500 2.69% 2,740 15:40 0 10:00
4/4/01 9:45 Buy 92.00 4/4/01 13:50 92.00 49 0 0.00% 1,900 11:20 (1,890) 10:30
4/4/01 13:50 Sell 92.00 4/4/01 15:55 91.85 25 150 0.16% 380 14:00 (500) 14:20
4/5/01 9:40 Buy 95.68 4/5/01 15:55 98.15 75 2,470 2.58% 3,040 15:25 (10) 9:40
4/6/01 9:40 Sell 97.30 4/6/01 11:55 98.24 27 (940) -0.97% 550 11:15 (940) 11:55
4/6/01 11:55 Buy 98.24 4/6/01 12:35 97.30 8 (940) -0.96% 1,660 12:05 (940) 12:35
4/6/01 12:35 Sell 97.30 4/6/01 15:55 97.67 40 (370) -0.38% 300 12:35 (1,960) 13:55

ACTIVE TRADER • September 2001 • www.activetradermag.com 44


A note on price data and dividends

A n overlooked aspect of testing a stock trading strategy is the effect of March 23 and Feb. 28 to March 30.
dividends. For example, IBM pays dividends on a quarterly basis, usu- Figure 1 is a five-minute chart with
ally on the “dividend payable dates” of March 10, June 10, Sept. 10 the noise channel superimposed, as well
and Dec. 10. On the “Ex-dividend dates” (approximately one month before as some of the buy and sell signals from
the payable date), the price of the stock is adjusted down by the value of the the Table 4 trade-by-trade summary.
dividend.
Thus, over the course of a year, IBM has a small downward price bias equal Breaking down the numbers
to the amount of the yearly dividend. If you were an owner of IBM, you would With respect to average winning and los-
receive those dividends in cash, making up for the small downward bias. ing trades, drawdowns and profit factor,
However, when developing and testing a system using historical stock data, the out-of-sample performance (Table 3)
prices are not adjusted for dividend payments. This creates a small distortion was better than the test sample perform-
in parameter selection and forward-adjusted results. ance (Tables 2a and 2b) The better per-
Because no dividends were paid in the data sample used for the test in this formance of the out-of-sample section
article, no adjustment needs to be made. However, if the intraday time peri- could have been coincidental, but it does
od fell on an ex-dividend date, an adjustment would have to be made to indicate that four weeks of test data was
avoid distortion. enough to capture the intraday price
dynamics of this stock.

Proper system testing

W hen testing any trading strategy, the important


point is how well it will perform on price data it has
not been optimized on — that is, out-of-sample
data. In short, without out-of-sample testing, it’s nothing
more than a hope and a prayer to believe that system per-
These parameter values were then applied to an out-of-
sample data period following the test segment (March 26 to
March 30). This walk-forward process was repeated by moving
the test data window forward one week, to Feb. 28 to March
30, and again finding the parameters values through optimiza-
formance in the future will be anywhere near the optimized tion on this new data. These optimized parameter values are
performance. then applied to the next out-of-sample five-minute intraday
For example, it’s possible to take a trading strategy with data window (April 2 to April 6). An important (but unspoken)
four independent variables, or parameters, and with hind- point in walk-forward testing is that if you cannot get good
sight, find the values for each of them that give the best results in the out-of-sample data segments, real-time system
(optimized) results on a specific historical period — say, the performance will be random.
last three years (using daily price data). Almost any period of historical prices can be curve fitted
However, these optimized parameter values have, in easily to give the false illusion of future profitability.
essence, been cherry-picked for this particular data period (a However, these performance measures in no way reflect how
process known as “curve-fitting”), and are unlikely to per- a system will perform on price data it has not been optimized
form as well on other historical test periods, or in actual trad- on. Only out-of-sample testing — that is, testing on price data
ing in the future. the system parameters were not originally derived from — can
A walk-forward testing procedure was applied to the noise determine if a system is robust and has a chance of perform-
channel breakout system as follows: Five-minute bars from a ing well in real trading.
period of four weeks from the start of the test period — Feb. Despite these facts, many market pundits still make the
21 to March 23 — were chosen and system parameter values unproven claim that statistics generated solely from optimized
were found through optimization on this intraday data seg- buy and sell trades in the test section (the initial period of
ment. In other words, the “best-performing” system parame- price data)have value in predicting whether or not the system
ters (e.g., number of days in lookback period, noise filter will perform well in the future. Nothing could be further from
value) were determined by testing a range of values for each. the truth. The only thing the statistics from the test section
At this stage of system development, the only thing indicated tell you is how well you have curve-fitted the data in the test
by the optimum values in the test portion is that the data has section. As a matter of fact, using optimization, it’s almost
been “curve-fitted” as best it can with this system. Without impossible not to get an excellent fit with great statistical
further testing on out-of-sample data, there is no way to tell if results.
the system will work in the future.

45 www.activetradermag.com • September 2001 • ACTIVE TRADER


The out-of-sample trade summary opening surprises. Overall, traded on dures and determine the appropriate
(Table 4) shows the system did better on IBM, the NCBS did a good job of mini- parameters for each. Every market has
short trades than it did on long trades. On mizing the whipsaw losses prevalent in subtle differences because the partici-
one hand, this could indicate a “negative” breakout trading systems and maximiz- pants vary from market to market. Also,
bias for the system. On the other hand, ing the profits from major intraday trend market activity can change over time.
given the current bear market environ- moves. Consequently, you should continue to
ment, the ability to cash in on the short perform walk-forward testing to deter-
side has value. There were no big winners Building on the results mine if there is a shift in the system’s
or big losers, indicating steady returns. To use this system in real-time trading, at effectiveness and whether better param-
Average wins were 2.4 times average least 10 additional test and out-of-sample eters have emerged.
losses in the out-of-sample section. windows should be examined to ensure Any trading method should be tested
Figure 1 shows how the system was the performance summarized here was before you risk capital on the technique.
able to efficiently capture intraday not the result of chance. Granted, there is a considerable amount
trends in IBM. Also, the system con- To determine if this approach can be of work involved, but without taking the
straint of not carrying positions used on other stocks or markets it would time to adequately research a technique,
overnight eliminated many negative be necessary to follow the same proce- the chances of success are low. 

FIGURE 1 NOISE CHANNEL AND TRADE SIGNALS


A few of the buy and sell signals generated by the noise channel breakout system are shown. The system successfully
captured intraday trends.

International Business Machine (IBM), five-minute


101

100

99
-1

98

97

96

95
1

94

93

3/26 11:20 12:15 1:10 2:05 3:00 3/27 11:15 12:10 1:05 2:00 2:55 3/28 11:10 12:05 1:00 1:55 2:50
Source: TradeStation by TradeStation Group

ACTIVE TRADER • September 2001 • www.activetradermag.com 46


TRADING Strategies

The long and short of it:


The Noise Channel Breakout
SYSTEM 2 BY DENNIS MEYERS, PH.D.

A lthough price breakouts are the basis for many


trading approaches, breakout systems are
plagued by false signals — when price initially
breaks out, triggering a buy or sell, but quickly
retraces, resulting in a losing trade.
To combat this problem, traders often apply filters to break-
out systems, delaying trade entry until the initial breakout has
Here we will use the NCBS, again applied to IBM five-
minute price bars, to see if some improvement can be made by
using different filters for long and short trades, respectively. To
compare the new version of the system to the previous one, the
following tests will use the same five-minute bar prices of IBM
from Feb. 21, 2001, to April 6, 2001.
First, we’ll review the basics of breakout systems in general
been confirmed by a price move of a certain size or duration in and the NCBS in particular.
the direction of the breakout.
“Better breakout trading: The noise channel breakout sys- NCBS refresher
tem“ (Active Trader, September 2001, p. 70) showed how a sim- The basic channel breakout system goes long on a move above
ple channel breakout system, with an additional noise filter to the highest high of the last n bars and goes short on a move
minimize whipsaws, could be developed to trade IBM on an below the lowest low of the last n bars. For example, a 40-day
intraday basis using five-minute bars. The noise filter delays channel breakout goes long when price moves above the high-
taking a breakout signal until the market provides some con- est high of the last 40 days and goes short when price falls
firmation the breakout is sustainable, thus avoiding false below the lowest low of the last 40 days.
breakouts. Breakout systems can be used on intraday price data, as well
One aspect the original noise channel breakout system as daily or weekly data. The NCBS is an intraday breakout sys-
(NCBS) did not address is whether to treat the long and short tem based on five-minute bars. Because intraday price action
sides of the market the same — that is, whether the filter can be very volatile, without some kind of filter the losses gen-
should be different for long and short trades, since uptrends erated by the random price movement (that is, whipsaws) can
and downtrends have different characteristics. completely overwhelm a trading system. To help eliminate

47 www.activetradermag.com • October 2001 • ACTIVE TRADER


Traders often use additional

rules or filters to prevent

being whipsawed by

breakout trading strategies.

Find out if using different WALK-FORWARD:


Proper system testing
filters for long and short

trades improves the

performance of an intraday
W hen testing any trading strategy, the
important point is how well it will
perform on data on which it has not
been optimized — that is, out-of-sample data.
If a certain system is first tested on a “sample”
piece of historical price data (say, daily bars
breakout strategy. from 1993 up to 1998), the system’s perform-
ance results have no implication outside this
sample data set; all you know is how well your
system parameters performed during this par-
ticular period.
To get an idea of how the system might actu-
ally perform, the system parameters used for
the sample data should be applied to different
“out-of-sample” price data (say, daily bars
from 1998 to the present). This “walk-forward
process” simulates the application of a system
to future data, as would occur in actual trad-
ing. In short, without out-of-sample testing,
such random movement, the NCBS adds a “noise filter,” designat- it’s nothing more than hope to believe that sys-
ed by the symbol f, to the basic channel breakout system. The three tem performance in the future will be any-
system parameters for the NCBS are: where near the optimized performance.
For example, it’s possible to take a trading
• nhi, which is the number of bars in the lookback period strategy with four independent variables, or
used to determine the highest high price (hhp). parameters, and with hindsight, find the values
• nlo, which is the number of bars in the lookback period for each of them that give the best (optimized)
used to determine the lowest low price (llp). results on a specific historical period — say, the
• f, which is the amount price must exceed the hhp or llp last three years (using daily price data).
to trigger a buy or sell. However, these optimized parameter values
have been, in essence, cherry-picked for this
The Noise Channel Breakout System 2 particular data period (a process known as
The Noise Channel Breakout System 2 (NCBS-2) uses different fil- “curve-fitting”), and are unlikely to perform as
ter values (f, from the original system) for the long and short sides well on other historical test periods, or in actu-
of the market. As a result, there are four system parameters for the al trading in the future. An important (but
NCBS-2: unspoken) point in walk-forward testing is that
if you cannot get good results in the out-of-
• nhi, which is the number of bars in the lookback period sample data segments, real-time system per-
used to determine the highest high price (hhp). formance will be random.
• nlo, which is the number of bars in the lookback period
used to determine the lowest low price (llp).

ACTIVE TRADER • October 2001 • www.activetradermag.com 48


• xoU, which is the amount price must exceed the hhp to
trigger a buy signal. Think of the symbols xoU and xoD as the “crossover Up”
• xoD, which is the amount price must fall below the llp to and “crossover Down” values.
trigger a sell signal. The logic behind modifying the filter values is because mar-
ket behavior associated with buys and sells is quite different,
TABLE 1 OPTIMUM PARAMETER VALUES FOR TEST DATA the noise channels associated with buys and sells should be
independent of each other. The NCBS-2 rules are:
Start date End date nhi xoU nlo xoD Buy rule: If price crosses above the highest high price of the
last nhi bars by an amount greater then or equal to xoU, then
2/21/01 3/23/01 8 1 4 1 buy at the market. In addition, when short, and when calculat-
ing the highest high price (hhp), the hhp can only stay the same
2/28/01 3/30/01 18 1.25 12 0.25 or go lower than its most recent value, it cannot go higher.
Sell rule: If price crosses below the lowest low price of last
nlo days by an amount of greater than or
equal to xoD, then sell at the market. In
FIGURE 1A TEST PERIOD 1 addition, when long and when calculat-
Performance summary for NCBS-2, IBM five-min. bars, Feb. 21 to March 23. ing the lowest low price (llp), the llp can
Statistics based upon buying and selling 1,000 shares of IBM. only stay the same or go higher than its
most recent value, it cannot go lower.
Performance summary: All trades Exit rule: Close any position five min-
Total net profit ($): 13,890 Open position P/L ($): 0 utes before the New York Stock Exchange
close (no trades are carried overnight).
Gross profit ($): 39,260 Gross loss ($): -25,370
Total no. of trades: 48 Percent profitable (%): 54 Testing the strategy
Number winning trades: 26 Number losing trades: 22 As in last month’s article, “walk-for-
Largest winning trade ($): 5,940 Largest losing trade ($): -2,060
ward” optimization will be used here.
The same data will also be used so we can
Average winning trade ($): 1,510 Average losing trade ($): -1,153.18 judge whether this new modification can
Ratio avg. win/avg. loss: 1.309 Avg. trade(win & loss) ($): 289.38 improve performance.
Max. consec. winners: 4 Max. consec. losers: 3 The walk-forward testing procedure
was applied as follows: A four-week peri-
Avg. no. bars in winners: 39 Avg. no. bars in losers: 21
od from the start of the IBM five-minute
Max intraday drawdown ($): -8,470 bar data from Feb. 21 through March 23
Profit factor: 1.547 Max. no. contracts held: 1 was chosen and system parameter values
were found through optimization on this
intraday data segment. (Optimization
FIGURE 1B TEST PERIOD 2 refers to the search for the parameter val-
ues that give the best results.) It should be
Performance summary for NCBS-2, IBM five-min. bars. Feb. 28 to March 30. noted that in this stage of system devel-
Statistics based upon buying and selling 1,000 shares of IBM. opment, the only thing indicated by the
optimum values that are found in the test
Performance summary: All trades portion is that the data has been curve fit-
Total net profit ($): 9,640 Open position P/L ($): 0 ted as best it can with this system.
Gross profit ($): 34,460 Gross loss ($): -24,820 Without further testing on out-of-sample
data, there is no way to tell if the system
Total no. of trades: 38 Percent profitable (%): 52.63
will work in the future.
Number winning trades: 20 Number losing trades: 18 The parameter values found were then
Largest winning trade ($): 5,350 Largest losing trade ($): -3,400 applied to an out-of-sample period, in
Average winning trade ($): 1,723 Average losing trade ($): -1,378.89 this case March 26 to March 30. This
process was repeated by moving the test
Ratio avg. win/avg. loss: 1.25 Avg. trade(win & loss) ($): 253.68
data window forward one week to Feb. 28
Max. consec. winners: 3 Max. consec. losers: 2 through March 30 and again finding the
Avg. no. bars in winners: 48 Avg. no. bars in losers: 28 parameters values through optimization
on this new data test window. The param-
Max. intraday drawdown ($): -10,030
eter values found were then applied to
Profit factor: 1.39 Max. no. contracts held: 1 the next out-of-sample data set, which in
Source: TradeStation by TradeStation Group Inc.
this case was April 2 to April 6. See “Walk-
forward: Proper system testing” for addi-

44 www.activetradermag.com • October 2001 • ACTIVE TRADER


tional information on optimization and walk-forward testing.
Of the four system parameters to find (nhi, nlo, xoU and Improved performance?
xoD), the “best” parameters are defined as those values that The optimum parameters in Table 1 show the first test data sec-
give the best net profits and best total winning bars/total los- tion produced the same optimum parameters as the original
ing bars ratio with the minimum drawdown and minimum NCBS. This can been seen by observing that both xoU and xoD
largest losing trades. In addition, the results should be stable are exactly the same and are equal to f of the original NCBS.
— e.g. the profits, wins and drawdowns should not change by The sample performance summaries in Figures 1a and 1b,
much as the parameters move by a small amount away from and the out-of-sample performance summary of Figure 2a,
their optimum values. Also, in choosing the best parameters, show the out-of-sample performance was better than the test
we considered only those parameter sets with maximum con- sample performance with respect to average winning and los-
secutive losses of four or less. ing trades, drawdowns and profit factor. This improved per-
formance in the out-of-sample section could have been due to
Results
Table 1 shows the optimum parameters
for the IBM five-minute data series. The FIGURE 2A TEST PERIOD 1
lookback periods refer to number of bars Combined walk-forward out-of-sample performance summary for NCBS-2,
and the filters values are given as dollar IBM five-min. bars, March 26 to April 6. Statistics based upon buying and
amounts. selling 1,000 shares of IBM.
Figures 1a and 1b) show the perform-
ance summary of the test windows using Performance summary: All trades
the optimum parameters shown in Table 1. Total net profit ($): 8,650 Open position P/L ($): 0
Figure 2a shows the combined per- Gross profit ($): 15,390 Gross loss ($): -6,740
formance summary of the two out-of-
Total no. of trades: 15 Percent profitable (%): 0.47
sample data segments from March 26 to
April 6 for NCBS-2. This performance Number winning trades: 7 Number losing trades: 8
represents what would have happened Largest winning trade ($): 4,000 Largest losing trade ($): -1,730
in real time if the parameters found in
Average winning trade ($): 2,198.57 Average losing trade ($): -842.50
the test sections (Table 1) were used.
Slippage and commissions are not Ratio avg. win/avg. loss: 2.61 Avg. trade(win & loss) ($): 576.67
included. For comparison, Figure 2b Max. consec. winners: 2 Max. consec. losers: 3
(bottom, right) shows the combined per- Avg. no. bars in winners: 57 Avg. no. bars in losers: 38
formance summary of the two out-of-
Max intraday drawdown ($): -5,660
sample data segments from March 26 to
April 6 for the original NCBS. Profit factor: 2.28 Max. no. contracts held: 1
Figure 3 shows a specialized percent-
age trade-by-trade summary from March FIGURE 2B TEST PERIOD 2
26 to April 6. Note that the trades from
March 26 to April 6 are the out-of-sample Combined walk-forward out-of-sample performance summary for NCBS, IBM
trades generated from the optimized five-min. bars, March 26 to April 6. Statistics based upon buying and selling
parameters from the two test sections of 1,000 shares of IBM.
Feb. 21 to March 23 and Feb. 28 to March
Performance summary: All trades
30. The in-sample trades are, by defini-
tion, curve-fit and are not of interest here. Total net profit ($): 8,390 Open position P/L ($): 0
In addition, for comparison with Gross profit ($): 14,460 Gross loss ($): -6,070
Figure 3, Figure 4 contains the special- Total no. of trades: 16 Percent profitable (%): 50
ized trade-by-trade summary from the
Number winning trades: 8 Number losing trades: 8
original NCBS for the same out-of-sam-
ple dates. Largest winning trade ($): 4,000 Largest losing trade ($): -1,350
Figure 5 is a five-minute chart of IBM Average winning trade ($): 1,807.50 Average losing trade ($): -758.75
with the NCBS-2 channels superimposed Ratio avg. win/avg. loss: 2.382 Avg. trade(win & loss) ($): 524.375
and some of the buy and sell signals
Max. consec. winners: 5 Max. consec. losers: 3
from the Figure 3 trade-by trade summa-
ry indicated on the charts. (All the sig- Avg. no. bars in winners: 54 Avg. no. bars in losers: 37
nals, as well as expanded performance Max. intraday drawdown ($): -4,480
statistics, can be found at www.activetra-
Profit factor: 2.382 Max. no. contracts held: 1
dermag.com.) Also included at the bot-
tom of the chart is the bar-by-bar profit Source: TradeStation by TradeStation Group Inc.
or loss of each trade.

ACTIVE TRADER • October 2001 • www.activetradermag.com 50


FIGURE 3 OUT-OF-SAMPLE TRADE BY TRADE SUMMARY
IBM five-min.; NCBS-2; Trade size = 1,000 shares; March 26 to April 6

Entry Entry Entry Exit Exit Exit Bars Trade Trade Trade Time Trade Time
date time price date time price in trade $ % Max Max
P&L P&L $Pft $DD
3/26/01 10:20 Sell 93.75 3/26/01 15:55 94.52 67 (770) (0.82) 0 10:20 (1,620) 10:35
3/27/01 10:15 Buy 95.59 3/27/01 15:55 99.59 68 4,000 4.18 4,300 15:50 0 10:15
3/28/01 9:40 Sell 97.92 3/28/01 15:55 94.50 75 3,420 3.49 3,420 12:00 (380) 9:40
3/29/01 10:05 Buy 96.05 3/29/01 15:05 94.90 60 (1,150) (1.20) 950 10:30 (1,160) 11:20
3/29/01 15:05 Sell 94.90 3/29/01 15:55 94.88 10 20 0.02 390 15:15 (500) 15:45
3/30/01 9:40 Buy 96.70 3/30/01 13:05 96.20 41 (500) (0.52) 800 11:55 (1,190) 10:00
3/30/01 13:05 Sell 96.20 3/30/01 15:55 96.25 34 (50) (0.05) 220 13:15 (840) 14:35
4/2/01 10:55 Sell 96.45 4/2/01 15:55 94.50 60 1,950 2.02 2,650 15:40 (1,250) 11:40
4/3/01 9:40 Sell 93.33 4/3/01 15:55 90.50 75 2,830 3.03 3,070 15:40 (670) 9:45
4/4/01 10:55 Buy 92.99 4/4/01 12:55 92.55 24 (440) (0.47) 910 11:20 (660) 11:00
4/4/01 12:55 Sell 92.55 4/4/01 15:55 91.85 36 700 0.76 930 14:00 (520) 13:20
4/5/01 9:40 Buy 95.68 4/5/01 15:55 98.15 75 2,470 2.58 3,040 15:25 (10) 9:40
4/6/01 9:40 Sell 97.30 4/6/01 12:00 98.75 28 (1,450) (1.49) 550 11:15 (1,450) 12:00
4/6/01 12:00 Buy 98.75 4/6/01 12:35 97.02 7 (1,730) (1.75) 1,150 12:05 (1,730) 12:35
4/6/01 12:35 Sell 97.02 4/6/01 15:55 97.67 40 (650) (0.67) 20 12:35 (2,240) 13:55
Total Average Average Average
8,650 0.61% 1,493 (948)
Source: Meyers Analytics

FIGURE 4 OUT-OF-SAMPLE TRADE BY TRADE SUMMARY


IBM five-min.; NCBS; Trade size = 1,000 shares; March 26 to April 6

Entry Entry Entry Exit Exit Exit Bars Trade Trade Trade Time Trade Time
date time price date time price in trade $ % Max Max
P&L P&L $Pft $DD
3/26/01 10:20 Sell 93.75 3/26/01 15:55 94.52 67 (770) (0.82) 0 10:20 (1,620) 10:35
3/27/01 10:15 Buy 95.59 3/27/01 15:55 99.59 68 4,000 4.18 4,300 15:50 0 10:15
3/28/01 9:40 Sell 97.92 3/28/01 15:55 94.50 75 3,420 3.49 3,420 12:00 (380) 9:40
3/29/01 10:05 Buy 96.05 3/29/01 15:05 94.90 60 (1,150) (1.20) 950 10:30 (1,160) 11:20
3/29/01 15:05 Sell 94.90 3/29/01 15:55 94.88 10 20 0.02 390 15:15 (500) 15:45
3/30/01 9:40 Buy 96.70 3/30/01 13:05 96.20 41 (500) (0.52) 800 11:55 (1,190) 10:00
3/30/01 13:05 Sell 96.20 3/30/01 15:55 96.25 34 (50) -0.05) 220 13:15 (840) 14:35
4/2/01 9:40 Buy 97.75 4/2/01 10:55 96.40 15 (1,350) -1.38) 350 10:05 (1,350) 10:55
4/2/01 10:55 Sell 96.40 4/2/01 15:55 94.50 60 1,900 1.97 2,600 15:40 (1,300) 11:40
4/3/01 10:00 Sell 93.00 4/3/01 15:55 90.50 71 2,500 2.69 2,740 15:40 0 10:00
4/4/01 9:45 Buy 92.00 4/4/01 13:50 92.00 49 0 0.00 1,900 11:20 (1,890) 10:30
4/4/01 13:50 Sell 92.00 4/4/01 15:55 91.85 25 150 0.16 380 14:00 (500) 14:20
4/5/01 9:40 Buy 95.68 4/5/01 15:55 98.15 75 2,470 2.58 3,040 15:25 (10) 9:40
4/6/01 9:40 Sell 97.30 4/6/01 11:55 98.24 27 (940) (0.97) 550 11:15 (940) 11:55
4/6/01 11:55 Buy 98.24 4/6/01 12:35 97.30 8 (940) (0.96) 1,660 12:05 (940) 12:35
4/6/01 12:35 Sell 97.30 4/6/01 15:55 97.67 40 (370) (0.38) 300 12:35 (1,960) 13:55
Total Average Average Average
8,390 0.55% 1,475 (911)
Source: Meyers Analytics

51 www.activetradermag.com • October 2001 • ACTIVE TRADER


FIGURE 5 NCBS-2 SIGNALS

Trade signals for the NCBS-2 are shown on a five-minute chart of IBM. The blue and red lines are the long and short
filter levels, respectively.

International Business Machine (IBM), five-minute


101
0
100

-1 99

98

97

96

95
-1

1
94
0
0

93

4,000

2,500

1,000

500

9:55 10:50 11:45 12:40 1:35 2:30 3/27 10:50 11:45 12:40 1:35 2:30 3/28 10:50 11:45 12:40 1:35 2:30

chance but does indicate that four weeks of test data were determine if the more complicated NCBS-2 offers any advan-
enough to capture the intraday price dynamics of IBM. tage in the trade-by-trade figures. There seems to be little
The performance summaries in Figures 2a and 2b show advantage: Both systems’ totals and averages are nearly the
there is very little difference between the NCBS and NCBS-2. same. The NCBS-2 had one less trade and slightly better num-
The less-complicated NCBS, while having a slightly lower net bers. However, the difference wasn’t enough to claim any
profit and average win/average loss ratio, has a smaller draw- superiority or to justify the added complication of another
down and a smaller largest losing trade. Comparison of optimization parameter.
Figures 2a and 2b favors the simpler NCBS. The NCBS-2 did very well in catching every major intraday
The out-of-sample trade-by-trade summary of Figure 3 trend of IBM. The charts show the system constraint of not car-
shows the system did better on short trades than on long rying positions overnight eliminated many negative opening
trades. This could indicate a negative bias for the system, or surprises. Overall, the system did a good job in minimizing the
perhaps, given the current bear market, this could be normal. losses resulting from the inevitable whipsaws that will occur in
Whatever the reason, this bias warrants further investigation. any trading system and maximizing the profits from the major
There were no big winners or big losers, indicating steady intraday trend moves of IBM.
returns. Average wins were 2.6 times average losses in the out-of- To use NCBS-2 in real time trading, the results from at least
sample section. Average trade run-ups were $1,493, average trade 10 to 20 more tests and out-of-sample periods would have to be
drawdowns were -$948 and the average trade net profit was $576. examined to make sure that the results above were not due to
It’s also instructive to compare Figure 3 with Figure 4 to pure chance. 

ACTIVE TRADER • October 2001 • www.activetradermag.com 52


ADVANCED Strategies

The multibar range BREAKOUT SYSTEM


Breakouts of price channels can be
profitable — if the volatility is there
and you’re on the right side of the
trade. This stop-and-reverse system
tries to capture intraday trends in
the S&P E-Mini contract by recognizing
differences in the characteristics
of up moves and down moves.

BY DENNIS MEYERS, PH.D.


FIGURE 1 TRADESTATION CODE FOR THE MULTIBAR
RANGE BREAKOUT SYSTEM

B
{Strategy: #MultiBarRangeBO}
reakout systems are popular when markets are
Input: n(45),bx(0.45),m(15),sx(0.45),XTime(1515);
volatile. Such systems typically identify support
vars: hhv1(h),llv1(l),hhv2(h),llv2(l),ii(0),xb(c),xs(c);
and resistance levels when price has been moving
in a range or channel, and enter trades when price
hhv1=h; llv1=l;
breaks out of either the up side or down side of a channel.
for ii=1 to n-1 begin
There are two simple ways to define support and resistance
if h[ii]>hhv1 then hhv1=h[ii]; if l[ii]<llv1 then llv1=l[ii];
levels for price channels. In both cases, it is first necessary to
end;
define a lookback period. The first way is to use the highest
value1=hhv1-llv1;
high and the lowest low of the lookback period. The second
way is to determine the range of each bar (high minus low) and
hhv2=h; llv2=l;
add that range (or a percentage of it) to, or subtract it from, the
for ii=1 to m-1 begin
current close.
if h[ii]>hhv2 then hhv2=h[ii]; if l[ii]<llv2 then llv2=l[ii];
In either case, the upper and lower boundaries represent the
end;
price channel. One advantage to the second method is it better
value2=hhv2-llv2;
reflects the volatility of the market — it will expand and con-
tract as the volatility changes.
xb= c + (Value1 * bx);
Breakout strategies require the market to be in a high-volatil-
xs= c - (Value2 * sx);
ity period; a trade will become profitable only if it continues to
move in the direction of the breakout. Volatility and emotion go
if time<XTime then begin
hand in hand. As volatility increases, traders have to cope with
if marketposition<=0 then Buy Next Bar xb stop;
more risk; hence, the more emotional the market becomes. This
if marketposition>=0 then Sell Short Next Bar xs stop;
is often reflected by the fact markets fall faster than they rise.
end;
In the following system, the channel is determined by using
the range of the price bars in the lookback period. A breakout
if XTime<>0 then SetExitOnClose;
above or below the channel’s resistance or support creates buy
or sell signals.

53 www.activetradermag.com • January 2004 • ACTIVE TRADER


However, the parameters for the buy signals will be TABLE 1 MULTIBAR RANGE BREAKOUT SYSTEM
different than those for the sell signals, because of the PERFORMANCE SUMMARY, JULY 7 TO AUG. 1, 2003
propensity for markets to fall faster than they rise. The
range for the last x bars will be defined as the highest The system triggered more short trades during the test, but pro-
high of the last x bars (including the current bar) minus duced profits on long trades, as well.
the lowest low of the last x bars (including the current
bar).
All trades Long trades Short trades
The buy price is determined by adding a percentage
of the range of the last n bars to the current close — the Total net profit $4,912.50 $1,450.00 $3,462.50
previously described volatility-adjusted technique. If Gross profit $6,637.50 $1,912.50 $4,725.00
the next bar’s price exceeds the buy price, the system Gross loss ($1,725.00) ($462.50) ($1,262.50)
issues a buy signal. The sell price is determined by
Profit factor 3.85 4.14 3.74
subtracting a percentage (a different percentage than
the buy percentage) of the range of the last m bars from Open position P/L $0.00 $0.00 $0.00
the current close. If the next bar’s price falls below the
sell price, the system issues a sell signal. Total number of trades 55 20 35
The resulting Multibar Channel Breakout system Percent profitable 58.18% 55.00% 60.00%
will trade the S&P 500 E-Mini futures on an intraday
Winning trades 32 11 21
basis using one-minute bars. The TradeStation Code is
shown in Figure 1 (opposite page). Losing trades 22 9 13
Even trades 1 0 1
Multibar Channel Breakout rules
This is a stop-and-reverse system, meaning it is always Avg. trade net profit $89.32 $72.50 $98.93
in the market: When a sell signal occurs, long trades Avg. winning trade $207.42 $173.86 $225.00
are exited and a short trade is entered; when a buy sig- Avg. losing trade ($78.41) ($51.39) ($97.12)
nal occurs, short trades are exited and a long trade is
Ratio avg. winning/
entered. These are the system’s parameters:
avg. losing 2.65 3.38 2.32
ES = E-Mini price; Largest winning trade $700.00 $362.50 $700.00
BRange = the price range over the last n bars; Largest losing trade ($300.00) ($125.00) ($300.00)
SRange = the price range over the last m bars;
bx = the percentage multiplier of the BRange for buy Largest winner as
signals; % of gross profit 10.55% 18.95% 14.81%
sx = the percentage multiplier of the SRange for sell Largest loser as
signals; % of gross loss 17.39% 27.03% 23.76%
c = the current price;
Net profit as
buyCh = c + bx*BRange;
% of largest loss 1,637.50% 1,160.00% 1,154.17%
sellCh = c - sx*SRange
Max. consecutive
where winning trades 5 5 4
n = The number of lookback bars (including the cur-
Max. consecutive
rent bar) for buy signals.
losing trades 5 2 3
m = The number of lookback bars (including the
current bar) for sell signals. Avg. bars in
total trades 134.96 45.8 185.91
Notice that not only are the percentage multipliers Avg. bars in
for long (bx) and short trades (sx) different, the look- winning trades 166.88 69.18 218.05
back periods the system references for buys (n) and
Avg. bars in
sells (m) are also different. The trade rules are simple:
losing trades 91.95 17.22 143.69
1. Buy rule: Buy the next bar at buyCh, stop. Max. drawdown
2. Sell rule: Sell the next bar at sellCh, stop. (intraday peak to valley) ($887.50) ($862.50) ($975.00)
3. Intraday bar exit rule: Exit the position on the
Max. drawdown
close (no overnight trades).
(trade close to trade close) ($300.00) ($175.00) ($400.00)
Although it may not be immediately obvious, this sys- Max. trade drawdown ($475.00) ($475.00) ($362.50)
tem avoids the opening gap whipsaw problem —
trades being triggered because of large gap openings Source: TradeStation

ACTIVE TRADER • January 2004 • www.activetradermag.com 54


TABLE 2 TRADE-BY-TRADE SUMMARY: JULY 7 TO AUG. 1, 2003
This list contains each trade in the test period. Overall, 58.18 percent of trades were profitable.
Entry Entry Entry Exit Exit Exit Bars Trade Trade Trade
date time price ($) date time price ($) in trade $P&L max$Pft Time max$DD Time
7/7/03 10:36 Sell 1,003.75 7/7/03 12:35 1,002.50 119 $62.50 $175.00 12:06 ($12.50) 10:36
7/7/03 12:35 Buy 1,002.50 7/7/03 12:52 1,001.50 17 ($50.00) $0.00 12:35 ($50.00) 12:38
7/7/03 12:52 Sell 1,001.50 7/7/03 13:01 1,002.75 9 ($62.50) $25.00 12:55 ($62.50) 13:01
7/7/03 13:01 Buy 1,002.75 7/7/03 13:24 1,002.50 23 ($12.50) $37.50 13:23 ($25.00) 13:08
7/7/03 13:24 Sell 1,002.50 7/7/03 15:15 1,002.75 111 ($12.50) $87.50 14:22 ($112.50) 13:47
7/8/03 9:51 Sell 1,002.00 7/8/03 14:09 1,004.25 258 ($112.50) $62.50 11:21 ($175.00) 10:19
7/8/03 14:09 Buy 1,004.25 7/8/03 15:15 1,007.50 66 $162.50 $187.50 14:48 ($62.50) 14:11
7/9/03 9:33 Sell 1,007.75 7/9/03 15:15 1,001.00 342 $337.50 $525.00 11:03 $0.00 9:33
7/10/03 8:58 Sell 992.50 7/10/03 15:15 988.75 377 $187.50 $512.50 13:48 ($62.50) 9:04
7/11/03 9:05 Sell 993.25 7/11/03 15:15 997.75 370 ($225.00) $62.50 9:08 ($337.50) 11:23
7/14/03 9:54 Sell 1,012.25 7/14/03 15:15 1,002.75 317 $475.00 $575.00 14:46 ($112.50) 10:01
7/15/03 9:05 Sell 1,002.75 7/15/03 15:15 1,000.75 370 $100.00 $362.50 14:05 ($275.00) 9:46
7/16/03 8:48 Sell 1,000.50 7/16/03 12:10 993.25 202 $362.50 $625.00 10:12 $0.00 8:48
7/16/03 12:10 Buy 993.25 7/16/03 12:22 992.25 12 ($50.00) $0.00 12:10 ($50.00) 12:18
7/16/03 12:22 Sell 992.25 7/16/03 15:15 995.25 173 ($150.00) $187.50 14:30 ($150.00) 15:10
7/17/03 9:04 Sell 988.25 7/17/03 11:01 986.25 117 $100.00 $275.00 9:39 $0.00 9:04
7/17/03 11:01 Buy 986.25 7/17/03 11:02 983.75 1 ($125.00) $0.00 11:01 ($125.00) 11:02
7/17/03 11:02 Sell 983.75 7/17/03 15:15 980.50 253 $162.50 $337.50 14:01 ($25.00) 11:06
7/18/03 8:49 Sell 986.00 7/18/03 11:27 985.25 158 $37.50 $287.50 9:20 ($25.00) 9:02
7/18/03 11:27 Buy 985.25 7/18/03 12:02 985.50 35 $12.50 $62.50 11:31 ($12.50) 11:27
7/18/03 12:02 Sell 985.50 7/18/03 13:01 985.50 59 $0.00 $50.00 12:19 ($25.00) 12:03
7/18/03 13:01 Buy 985.50 7/18/03 14:35 991.50 94 $300.00 $375.00 14:11 $0.00 13:01
7/18/03 14:35 Sell 991.50 7/18/03 15:15 990.00 40 $75.00 $75.00 14:53 ($87.50) 14:41
7/21/03 8:32 Sell 988.00 7/21/03 15:15 978.25 403 $487.50 $700.00 14:10 ($12.50) 8:32
7/22/03 9:20 Sell 976.75 7/22/03 10:01 978.50 41 ($87.50) $112.50 9:50 ($87.50) 10:01
7/22/03 10:01 Buy 978.50 7/22/03 11:37 985.75 96 $362.50 $500.00 11:10 ($75.00) 10:18
7/22/03 11:37 Sell 985.75 7/22/03 14:37 987.25 180 ($75.00) $237.50 12:37 ($150.00) 14:02
7/22/03 14:37 Buy 987.25 7/22/03 15:15 986.75 38 ($25.00) $25.00 14:38 ($87.50) 14:49
7/23/03 8:35 Sell 986.25 7/23/03 12:51 984.75 256 $75.00 $412.50 11:16 $0.00 8:35
7/23/03 12:51 Buy 984.75 7/23/03 13:38 986.50 47 $87.50 $187.50 13:35 ($37.50) 12:55
7/23/03 13:38 Sell 986.50 7/23/03 14:27 986.75 49 ($12.50) $100.00 14:14 ($50.00) 13:39
7/23/03 14:27 Buy 986.75 7/23/03 15:15 987.75 48 $50.00 $62.50 15:13 ($62.50) 14:35
7/24/03 9:21 Sell 995.50 7/24/03 12:45 993.75 204 $87.50 $162.50 9:50 ($50.00) 10:27
7/24/03 12:45 Buy 993.75 7/24/03 13:10 994.25 25 $25.00 $62.50 12:57 $0.00 12:45
7/24/03 13:10 Sell 994.25 7/24/03 15:15 980.25 125 $700.00 $762.50 14:57 ($25.00) 13:11
7/25/03 8:43 Buy 982.50 7/25/03 9:01 983.25 18 $37.50 $150.00 9:00 ($37.50) 8:45
7/25/03 9:01 Sell 983.25 7/25/03 13:01 989.25 240 ($300.00) $375.00 9:52 ($337.50) 12:33
7/25/03 13:01 Buy 989.25 7/25/03 15:08 996.00 127 $337.50 $412.50 14:54 ($87.50) 13:25
7/25/03 15:08 Sell 996.00 7/25/03 15:15 997.00 7 ($50.00) $0.00 15:08 ($62.50) 15:12
7/28/03 8:33 Sell 995.50 7/28/03 12:37 996.50 244 ($50.00) $175.00 8:52 ($187.50) 10:29
7/28/03 12:37 Buy 996.50 7/28/03 12:57 996.25 20 ($12.50) $100.00 12:55 ($12.50) 12:37
7/28/03 12:57 Sell 996.25 7/28/03 15:15 993.50 138 $137.50 $187.50 14:42 ($87.50) 14:04
7/29/03 9:01 Sell 991.25 7/29/03 10:34 986.75 93 $225.00 $450.00 9:34 $0.00 9:01
7/29/03 10:34 Buy 986.75 7/29/03 11:28 992.00 54 $262.50 $450.00 10:58 ($12.50) 10:34
7/29/03 11:28 Sell 992.00 7/29/03 15:15 989.00 227 $150.00 $300.00 14:17 ($250.00) 12:33
7/30/03 8:34 Sell 990.00 7/30/03 11:38 989.25 184 $37.50 $275.00 10:32 ($37.50) 9:56
7/30/03 11:38 Buy 989.25 7/30/03 11:56 988.00 18 ($62.50) $0.00 11:38 ($62.50) 11:52
7/30/03 11:56 Sell 988.00 7/30/03 13:02 988.75 66 ($37.50) $62.50 12:19 ($50.00) 12:01
7/30/03 13:02 Buy 988.75 7/30/03 13:04 987.00 2 ($87.50) $0.00 13:02 ($87.50) 13:04
7/30/03 13:04 Sell 987.00 7/30/03 15:15 986.25 131 $37.50 $125.00 13:14 ($25.00) 14:21
7/31/03 9:00 Buy 995.75 7/31/03 11:20 1,001.25 140 $275.00 $387.50 10:14 ($400.00) 9:08
7/31/03 11:20 Sell 1,001.25 7/31/03 13:07 1,003.00 107 ($87.50) $37.50 11:40 ($112.50) 12:14
7/31/03 13:07 Buy 1,003.00 7/31/03 13:22 1,002.25 15 ($37.50) $12.50 13:19 ($50.00) 13:16
7/31/03 13:22 Sell 1,002.25 7/31/03 15:15 988.50 113 $687.50 $725.00 14:56 ($12.50) 13:22
8/1/03 8:46 Sell 983.50 8/1/03 15:15 979.50 389 $200.00 $300.00 9:35 ($125.00) 8:51
Source: Meyers Analytics, LLC

55 www.activetradermag.com • January 2004 • ACTIVE TRADER


FIGURE 2 RIDING THE TREND
During this period, the system caught one intraday uptrend, one intraday downtrend, and produced small losses on two
signals when the market was flat.

September 2003 S&P E-Mini futures (ESU03), one-minute


1,010

Short
Short 1,005

Buy 1,000

995

990
Buy Short

End 985
End of day
of day exit
exit 980

600
400
200
0
-200

7/30 9:11 9:33 9:55 10:17 10:39 11:01 11:23 11:45 12:07 12:29 12:51 13:13 13:35 13:57 14:19 14:41 8/1
Source: TradeStation

that quickly reverse and stop out the position. With this sys- difficult to sustain more than a handful of consecutive losses,
tem, if there is a gap on the opening bar, the buy and sell ranges we eliminated all cases that had more than five losing trades in
are expanded and no trades are made until the buy and sell a row. Of the remaining test results, we chose the one that had
ranges contract or the price breaks the expanded ranges. the highest total net profit and the lowest drawdown. The opti-
Breaking the expanded ranges takes time and avoids the open- mization procedure produced the following system parame-
ing gap whipsaw. ters:

Testing n = 45;
The system was tested from July 7 through Aug. 1, 2003, using bx = 0.45;
September 2003 E-Mini futures (ESU03) one-minute bars. A m = 15;
wide range of parameter values was tested to find the optimal sx = 0.45;
ones for the system. The parameter ranges tested for the initial
optimization test were: Table 1 (p. 43) shows the performance summary for the four-
week test period (slippage and commissions not included).
n =10 to 50 in steps of 5; Table 2 (opposite page) is a trade-by-trade summary of all the
bx = 0.4 to 1 in steps of 0.05; trades. The average net profit per trade was $89 — well above
m = 10 to 50 in steps of 5; slippage and commissions for a typical S&P E-Mini trade. The
sx = 0.4 to 1 in steps of 0.05; largest losing trade was $300, and the biggest intraday draw-
down was $887. These losses are small compared to the total
After the initial test, we had to choose one set of parameters net profit of $4,912.
that produced the most realistic results. To avoid curve fitting, Figures 2 and 3 are one-minute charts of the S&P E-Mini
we eliminated all results that had profit factors (gross profit that span July 31 to Aug. 1. The Multibar Range Breakout chan-
divided by gross loss) greater than 4.0, since such performance nels are superimposed on the price series, and all the buy and
was unlikely to be duplicated in the future. Also, because it is sell signals are marked. Finally, the bottoms of Figures 2 and 3

ACTIVE TRADER • January 2004 • www.activetradermag.com 56


FIGURE 3 ONE DAY, ONE TRADE
If no signal in the opposite direction is triggered, the system will stay in the same direction the entire day. All trades
are exited at the close — no positions are held overnight.

September 2003 S&P E-Mini futures (ESU03), one-minute 1,002


1,000
998
996
994
992
990
Sell 988
986
End of
day exit 984
982
980
978
End of 976
day exit

600

400

200

14:19 14:41 8/1 9:02 9:24 9:46 10:08 10:30 10:52 11:14 11:36 11:58 12:20 12:42 13:04 13:26 13:48 14:10 14:32 14:54

Source: TradeStation

FIGURE 4 DAILY PERSPECTIVES


include the bar-by-bar profit or loss of
each trade. The daily chart of the test period shows the system was able to profit on
Figure 4 is a daily chart of the S&P E- both sides of the market when conditions shifted from uptrend to downtrend
Mini futures from July 7 to Aug. 1, and to consolidation.
shows the market moved up, down and
sideways during this period. The system September 2003 S&P E-Mini futures (ESU03), daily 1,015
was able to produce profits on both the
long and short sides of the market, and 1.010
aside from a streak of five losing trades
near the outset of the test period, never 1,005
had more than three consecutive losses.
The Multibar Channel Breakout sys- 1,000
tem’s positive performance warrants
further investigation. If you consider 995
following this system in real-time, pay
close attention to how the real-time sta- 990
tistics compare to the hypothetical num-
bers shown here. If the numbers begin to 985
deviate, another review of the system
parameters are in order.  980

975
Individual articles can be purchased and
7 14 21 28
downloaded from www.activetradermag.com/
purchase_articles.htm. Source: TradeStation

57 www.activetradermag.com • January 2004 • ACTIVE TRADER


The TRADING Systems Lab

EQUITY CURVE
600,000
DeMark variation
500,000
Markets: Stocks, stock index futures, index stocks
(SPDRs, DIAs, QQQs), futures and currencies
400,000

Account balance ($)


System logic:
This system is based on a simple pattern, named TD
Carrie, described by Tom DeMark in his book New 300,000
Market Timing Techniques (John Wiley & Sons, 1997).
It trades a move above or below the true high (the
200,000
highest of one bar’s high and the previous bar’s
close) or the true low (the lowest of one bar’s low
and the previous bar’s close) of the bar four days 100,000
prior to the current (active) bar. However, for the
breakout to be valid it must be qualified by a few cri-
teria. (The following rules are described in terms of 0
a long trade; reverse for short trades.) 11/12/91 11/12/92 11/12/93 11/12/94 11/12/95 11/12/96 11/12/97 11/12/98 11/12/99 11/12/00
First, to identify a strongly trending market, the
true high of four days ago must be higher than the high five days ago. requiring the breakout to take place intraday, we enter into an
If this requirement is not met, it’s still possible to get an entry signal if orderly market instead of a highly volatile one.
the market has made a correction counter to the direction of an even-
tual trade (i.e., a downward correction in the case of a long trade). In Rules:
an uptrend this correction is identified by the highs of either two or 1. Prepare to go long today if
three days ago being lower than the true high of four days ago. a. the true high from four days ago is higher than the high from
Second, the close of the bar prior to the anticipated breakout either two, three or five days ago, and
needs to be lower then the previous bar’s close. This is to ensure that b. yesterday’s close is lower than the close two days ago, and
most traders still have a short-term bearish outlook prior to the c. today’s open is lower than the high four days ago.
upside breakout. That will increase the force of the up move as the 2. Prepare to go short today if
traders are caught on the wrong side of the market and scramble to a. the true low from four days ago is lower than the low from
get out of the market. either two, three or five days ago, and
Finally, the breakout must take place intraday and exceed the true b. yesterday’s close is higher than the close two days ago, and
high of four days ago by a sufficient amount. That the trade needs c. today’s open is higher than the low four days ago
to take place intraday means, for a valid upside breakout, the open- 3. Go long today with a stop order at the true high of four days
ing price of the day for the breakout must be lower than the true ago, plus 0.1 percent.
high of four days ago. This is to avoid entering into too strong an 4. Go short today with a stop order at the true low of four days
opening, which often marks the end of the current trend. Also, by ago, minus 0.1 percent.
5. Risk 2 percent of available equity per trade.
SAMPLE TRADES 6. Exit all trades with a loss if the market moves
against the position by 4 percent or more.
Amgen (AMGN), daily Sell 71.00 7. Exit all trades with a profit if the market
LX#3 LX moves in favor of the position by 12 percent or
69.00 more.
Buy 8. Exit all trades after five days, counting the
67.00
day for the entry as day one, and no matter how
65.00 late in the day the trade was made (i.e., a trade
LX#3
LX#3 Buy executed at 2:50 p.m. on Monday would be exit-
63.00 ed Friday the same week).
Sell
62.00
Sell
Buy
Test period: November 1991 to June 2001
60.00
Test data: Daily stock prices for the 30 highest
58.00
SX capitalized stocks in the Nasdaq 100 (excluding
Buy
Buy 56.00 Intel and Microsoft, which are also part of the
55.00 Dow Jones Industrial Average). $10 commission
SX#3 deducted per trade.
53.00
Starting equity: $100,000 (nominal)
51.00
2 9 16 23 30 7 14 21 28 4 11 Buy-and-hold stats:
April May June
Source: TradeStation by TradeStation Group Inc.
DJIA: Total return – 254 percent; Max DD – 22.5

58 www.activetradermag.com • September 2001 • ACTIVE TRADER


DRAWDOWN CURVE
11/12/91 11/12/92 11/12/93 11/12/94 11/12/95 11/12/96 11/12/97 11/12/98 11/12/99 11/12/00
0.00%

go with the entry strategies. We therefore arbitrarily


-5.00%
attached a 4-percent stop-loss and a 12-percent prof-
it-exit target, plus a time-based stop that exits all
-10.00%
trades after five days, no matter what. (All these
stops are completely un-optimized, which means a
-15.00%
little optimization should increase performance con-
siderably.)
-20.00%
Also, note the system operates with no trend filter,
such as a long-term moving average. Such filters,
-25.00%
which allow only those trades that are in the direction
of the underlying trend, also improve performance.
-30.00%
Finally, note that many of the stocks traded in this
example weren’t tradable until a few years ago,
-35.00%
which explains the initial large drawdown and
exceptionally long flat period. Had we been able to
-40.00%
test the same 30 stocks throughout the entire period,
percent (current); Longest flat – 18 months (current). it’s highly likely performance would have improved considerably.
S&P 500: Total return – 216 percent; Max DD – 30.4 percent (cur-
rent); Longest flat – 15 months (current).
Nasdaq: Total return – 519 percent; Max DD – 72 percent (current); ROLLING TIME WINDOW RETURN ANALYSIS
Longest flat – 15 months (current).
Cumulative 12 24 36 48 60
months months months months months
System analysis
Most recent: 4.27% 34.67% 81.32% 157.51% 336.03%
In DeMark’s original work, the amount by which the price had to
Average: 21.73% 59.85% 115.57% 185.15% 254.23%
clear the breakout level was set to the smallest price increment for
Best: 87.23% 206.04% 294.97% 393.04% 494.23%
the market in question. In this version, this is changed to one-tenth
Worst: -26.11% -25.61% -21.38% -15.55% 12.15%
of a percent to make the system consistent across all markets. This
means that for a stock that trades around $50, this amount is about
St. dev.: 29.26% 61.15% 95.27% 125.03% 137.24%
five cents; for a stock that trades around $100, it comes out to Annualized 12 24 36 48 60
approximately 10 cents. months months months months months
DeMark did not suggest any exit strategies or stop-loss levels to Most recent: 4.27% 16.05% 21.94% 26.68% 34.25%
Average: 21.73% 26.43% 29.18% 29.95% 28.78%
STRATEGY SUMMARY Best: 87.23% 74.94% 58.07% 49.01% 42.82%
Profitability Trade statistics Worst: -26.11% -13.75% -7.70% -4.14% 2.32%
St. dev: 28.26% 26.94% 24.99% 22.48% 18.86%
End. equity ($): 415,573 No. trades: 3,529
Total return (%): 316 Avg. trade ($): 158 LEGEND: Cumulative returns — Most recent: most recent return from start to
Avg. annual ret. (%): 16.03 Avg. DIT: 3.0 end of the respective periods • Average: the average of all cumulative returns from
start to end of the respective periods • Best: the best of all cumulative returns from
Profit factor: 1.13 Avg. win/loss ($): 1,150 (1,393) start to end of the respective periods • Worst: the worst of all cumulative returns
Avg. tied cap (%): 58 Lrg. win/loss ($): 12,674 (8,131) from start to end of the respective periods • St. dev: the standard deviation of all
cumulative returns from start to end of the respective periods
Win. months (%): 53 Win. trades (%): 39.4
Annualized returns — The ending equity as a result of the cumulative returns,
Drawdown TIM (%): 97 /15.1 raised by 1/n, where n is the respective period in number of years
Max DD (%): 37.5 Tr./Mark./Year: 12.3
Send Active Trader your systems
Longest flat (m): 57.2 Tr./Month: 30.7 If you have a trading system or idea you’d like to see tested,
LEGEND: End. equity ($) — equity at the end of test period • Total return
send it to us at the Trading System Lab. We’ll test it on a
(%) — total percentage return over test period • Avg. annual ret. (%) —
portfolio of stocks or futures (for now, maximum 30 markets,
average continuously compounded annual return • Profit factor — gross
using daily data starting Jan. 1, 1990), using true portfolio
profit/gross loss • Avg. tied cap (%) — average percent of total available cap-
analysis/optimization.
ital tied up in open positions • Win. months (%) — percentage profitable
Most system-testing software only allows you to test one
months over test period • Max DD (%) — maximum drop in equity •
market at a time. Our system-testing technique lets all mar-
Longest flat — longest period, in months, spent between two equity highs •
kets share the same account and is based on the interaction
No. trades — number of trades • Avg. trade ($) — amount won or lost by
within the portfolio as a whole.
the average trade • Avg. DIT— average days in trade • Avg. win/loss ($)
Start by e-mailing system logic (in TradeStation’s
— average wining and losing trade, respectively • Lrg. win/loss ($) —
EasyLanguage or in an Excel spreadsheet) and a short descrip-
largest wining and losing trade, respectively • Win. trades (%) — percent
tion to editorial@activetradermag.com, and we’ll get back to
winning trades • TIM (%) — amount of time there is at least one open posi-
you.
tion for entire portfolio, and each market, respectively • Tr./Mark./Year —
Note: Each system must have a clearly defined stop-loss
trades per market per year • Tr./Month — trades per month for all markets
level and a suggested optimal amount to risk per trade.

Disclaimer: The Trading System Lab is intended for educational purposes only to provide a perspective on different market concepts. It is not meant to recommend or
promote any trading system or approach. Traders are advised to do their own research and testing to determine the validity of a trading idea. Past performance does not
guarantee future results; historical testing may not reflect a system’s behavior in real-time trading.

ACTIVE TRADER • September 2001 • www.activetradermag.com 59


EQUITY CURVE
$300,000
Dynamic
breakout system $250,000

Markets: Any market with a propensity to $200,000

Account balance ($)


trend.
$150,000
System logic: This system enters a long (short)
trade if the last closing price is above (below) the
$100,000
highest high (lowest low) of the lookback peri-
od. The word “dynamic” refers to the fact that
the lookback period will change based on the $50,000
volatility of the market.
This is similar to the system tested in the $0
January Trading System Lab (p. 50), when 3/26/93 3/26/94 3/26/95 3/26/96 3/26/97 3/26/98 3/26/99 3/26/00 3/26/01 3/26/02
Bollinger Bands were used to trigger trades.
Since the Bollinger Bands moved farther away
from price as the volatility of the market increased, the higher Test data: Daily prices for 30 of the most widely traded
the volatility, the more difficult it was to enter and exit trades. Nasdaq 100 stocks. $10 per trade deducted for slippage and
This month’s system functions in a related manner. The high- commission.
er the volatility, the longer the lookback period will be. Because
buys and sells will be based on price making a new high or low Starting equity: $100,000 (nominal).
for the lookback period, the longer the lookback period, the
more difficult it will be to enter or exit a trade. Buy-and-hold stats:
In this case, the lookback period can range from 20 to 60 days Total Maximum Longest
and will change daily depending on the level of volatility. Index return drawdown flat period
(Volatility reading is the daily standard deviation of the closing
price over the last 30 days.) DJIA 146% 39% (current) 33 months (current)
You can read more about the logic of the system in the S&P 500 100% 51% (current) 30 months (current)
Futures System Lab (p. 70), where we have tested it on 15 dif- Nasdaq 180% 83% (current) 30 months (current)
ferent commodity futures markets.
Test results: Although the system has fared no better than
Rules: buy-and-hold over the life of the test period, it has fared much
1. Go long on the open if yester-
day’s close is higher than the high- SAMPLE TRADES
est high of the lookback period.
2. Exit by reversing the position. Cisco (CSCO), daily
17.00

Reverse the rules for short trades. Buy 16.00

Money management: 15.00


Sell
1. Risk 2 percent of available Buy
14.00
equity per market traded.
2. The number of shares to trade
13.00
was calculated with the following Sell
formula:
12.00

ST = AC * PR / Dist
where 11.00
ST = Shares to trade
AC = Available capital 10.00
PR = Percent risked
Dist = Distance between the entry
price and the exit price on the day 9.00
of entry.

Test period:
May June July August September October
April 1993 to October 2002.
Source: Omega Research ProSuite

60 www.activetradermag.com • February 2003 • ACTIVE TRADER


DRAWDOWN CURVE
3/26/93 3/26/94 3/26/95 3/26/96 3/26/97 3/26/98 3/26/99 3/26/00 3/26/01 3/26/02
0% maximum open profit of each trade. Overall,
this would translate into more profitable
-5%
months, a smoother equity curve and a high-
-10% er average annual return.
-15% Another way to improve this system could
be to use different lookback periods for long
-20%
and short trades. Most likely, the lookback
-25% period for the short side would be shorter
-30%
than that for the long side. As with all sys-
tems tested in the Trading Systems Lab, this
-35% one is totally unoptimized. As a result, you
-40% should be able to increase profits consider-
ably by experimenting with different settings
-45%
for the indicators and adding a few ideas of
-50% your own.

better than a Nasdaq 100 buy-and-hold strategy over the last 30 ROLLING TIME WINDOW RETURN ANALYSIS
months (a 43 percent drawdown compared to the Nasdaq’s 83
percent). Cumulative 12 24 36 48 60
months months months months months
It is quite difficult to succeed over the long term with a sys-
tem that sells short because of the inherent upside bias of the Most recent: 8.69% -19.18% 15.28% 63.22% 26.23%
stock market, and the high volatility associated with bear mar- Average: 8.77% 20.27% 34.69% 46.45% 58.78%
kets. For this system, the high volatility makes it difficult to Best: 104.96% 106.03% 131.24% 148.03% 186.41%
enter potential winning short trades and exit losing trades.
Worst: -27.93% -38.42% -14.94% 5.25% 4.40%
The best way to improve the results for this system would
probably be to reverse the logic for the exit so that it would be St. dev.: 27.06% 30.46% 28.77% 33.48% 40.51%
easier to exit during times of high volatility. This would most
Annualized 12 24 36 48 60
likely result in more and smaller losing trades, but also in larg- months months months months months
er winning trades because the system would exit closer to the
Most recent: 8.69% -10.10% 4.85% 13.03% 4.77%
STRATEGY SUMMARY Average: 8.77% 9.67% 10.44% 10.01% 9.69%
Best: 104.96% 43.54% 32.24% 25.49% 23.42%
Profitability Trade statistics
Worst: -27.93% -21.53% -5.25% 1.29% 0.86%
End. equity ($): 164,849 No. trades: 1,133
St. dev.: 27.06% 14.22% 8.79% 7.49% 7.04%
Total return (%): 65 Avg. trade ($): 57
Avg. annual ret. (%): 5.36 Avg. DIT: 63.3 LEGEND: Cumulative returns — Most recent: most recent return from start to
end of the respective periods • Average: the average of all cumulative returns
Profit factor: 1.15 Avg. win/loss ($): 1,053 (435) from start to end of the respective periods • Best: the best of all cumulative
Avg. tied cap (%): 74 Lrg. win/loss ($): 37,484 (2,209) returns from start to end of the respective periods • Worst: the worst of all cumu-
lative returns from start to end of the respective periods • St. dev.: the standard
Win. months (%): 50 Win. trades (%): 34.1
deviation of all cumulative returns from start to end of the respective periods
Drawdown TIM (%): 100 92.8 Annualized returns — The ending equity as a result of the cumulative returns,
Max. DD (%): 43.1 Tr./Mark./Year: 3.9 raised by 1/n, where n is the respective period in number of years
Longest flat (m): 31.3 Tr./Month: 9.9
Send Active Trader your systems
If you have a trading system or idea you’d like tested, send it to
LEGEND: End. equity ($) — equity at the end of test period • Total return
(%) — total percentage return over test period • Avg. annual ret. (%) — us at the Trading System Lab. We’ll test it on a portfolio of
average continuously compounded annual return • Profit factor — gross stocks or futures (for now, maximum 60 markets, using the last
profit/gross loss • Avg. tied cap (%) — average percent of total available cap- 2,500 trading days), using true portfolio analysis/optimization.
ital tied up in open positions • Win. months (%) — percentage profitable Most system-testing software only allows you to test one mar-
months over test period • Max. DD (%) — maximum drop in equity • ket at a time. Our system-testing technique lets all markets
Longest flat — longest period, in months, spent between two equity highs • share the same account and is based on the interaction within
No. trades — number of trades • Avg. trade ($) — amount won or lost by the portfolio as a whole.
the average trade • Avg. DIT— average days in trade • Avg. win/loss ($) Start by e-mailing system logic (in TradeStation’s
— average winning and losing trade, respectively • Lrg. win/loss ($) — EasyLanguage or in an Excel spreadsheet) and a short description
largest winning and losing trade, respectively • Win. trades (%) — percent to editorial@activetradermag.com, and we’ll get back to you.
winning trades • TIM (%) — amount of time there is at least one open posi- Note: Each system must have a clearly defined stop-loss level
tion for entire portfolio, and each market, respectively • Tr./Mark./Year — and a suggested optimal amount to risk per trade.
trades per market per year • Tr./Month — trades per month for all markets
Disclaimer: The Trading System Lab is intended for educational purposes only to provide a perspective on different market concepts. It is not meant to recommend or
promote any trading system or approach. Traders are advised to do their own research and testing to determine the validity of a trading idea. Past performance does not
guarantee future results; historical testing may not reflect a system’s behavior in real-time trading.

ACTIVE TRADER • February 2003 • www.activetradermag.com 61


FUTURES & OPTIONS
Trading System Lab
EQUITY CURVE
Dynamic $160,000

$140,000
breakout system $120,000

Account balance ($)


Markets: Any markets with a propensity to trend. $100,000

$80,000
System logic: This is the same system tested on 30
Nasdaq stocks (p. 60, where you can read more about $60,000
the system’s logic).
$40,000
The system is based on the Donchian breakout sys-
tem, which enters a trade as soon as the market trades $20,000
above or below the highest or lowest price of the last
$0
four weeks (approximately 20 days). 3/26/93 3/26/94 3/26/95 3/26/96 3/26/97 3/26/98 3/26/99 3/26/00 3/26/01 3/26/02
The Donchian breakout system was invented by
Richard Donchian in the 1970s and refined by Richard Dennis in least some of the markets traded should be in strong trends.
the 1980s. The more popular it became, the more other traders These markets should be able to produce profits large enough
modified the system by varying the lookback periods, applying to make up for the whipsaw losses produced in the other mar-
other types of filters and attaching various money management kets plus enough additional profits to make trading worth-
rules. while.
The dynamic breakout system is a modified version of the
Donchian system that alters the lookback period between 20 Rules:
and 60 days depending on the volatility of the market. This sys- 1. Go long on the open if yesterday’s close is higher than
tem and the volatility breakout system used in the January the highest high of the lookback period.
Trading System Lab are likely the most commonly used strate- 2. Exit by reversing the position.
gies of all time, particularly in the commodity futures market.
This is a result of commodity futures markets’ historical ten- Reverse the rules for short trades.
dency to trend. The idea is that capturing a strong trending
move should more than make up for a large number of small Money management:
losing trades produced during times of consolidation and stag- 1. Risk 2 percent of available equity per trade.
nant prices. Applying the system to many different futures mar- 2. The number of contracts to trade was calculated with the
kets should result in a steady profit, as it is highly likely that at following formula:

CT = AC * PR / Dist
SAMPLE TRADES
where
Oats (O), daily
CT = Contracts to trade
210.00 AC = Available capital
Sell 200.00
PR = Percent risked
Dist = Distance between the
190.00 entry price and the exit price on the
180.00 day of entry.

170.00 Test period: April 1993 to October


2002.
160.00

150.00 Test data: Daily prices for 15 com-


modity futures markets: cocoa, cof-
140.00 fee, corn, cotton, feeder cattle, lum-
ber, oats, orange juice, pork bellies,
130.00 soybeans, soy meal, soy oil, rough
Buy rice, sugar and wheat.
120.00
Starting equity: $100,000 (nomi-
nal); $50 deducted for slippage and
110.00 commission per contract traded.

May June July August September October Test results: It is fair to say this sys-
Source: Omega Research ProSuite tem did not work very well from
1996 to 1999. However, since 1999 it

62 www.activetradermag.com • February 2003 • ACTIVE TRADER


DRAWDOWN CURVE
3/26/93 3/26/94 3/26/95 3/26/96 3/26/97 3/26/98 3/26/99 3/26/00 3/26/01 3/26/02
0%

-5%
However, as the performance summary shows,
-10% this system is potentially poised to launch itself into
a new period of prosperity. This shows the market
-15% works in cycles, and just when you’re about to
throw in the towel, things often take a turn for the
-20%
better.
Aside from the various ways of improving this
-25%
strategy suggested on p. 60, the best way to opti-
-30% mize it for the futures markets is to trade it in many
markets. Trading a large number of markets is pos-
-35% sible thanks to the relatively low margin require-
ments of the futures markets (often only 5 to10 per-
has gained ground little by little, slowly increasing its average cent of the total contract value).
annual return. For example, the average return for the last
three years has been 6.4 percent. However, the return for the
past 12 months has been 9.97 percent, and the return is close to
ROLLING TIME WINDOW RETURN ANALYSIS
30 percent for the last six months.
Granted, there still is a long way to go before the system can Cumulative 12 24 36 48 60
months months months months months
find its way out of a drawdown that has lasted for almost
seven years, but the recent upward trend confirms our findings Most recent: 9.97% 17.00% 20.46% 0.25% 2.89%
from last month that the long-term trend-following strategy is Average: 2.53% 4.64% 4.24% 2.94% 2.18%
ready to stage a comeback as a profitable trading system. Best: 44.61% 47.38% 47.95% 50.34% 49.68%
In the late 1990s, many analysts claimed that long-term Worst: -20.11% -19.85% -25.71% -28.62% -24.67%
trend-following systems would no longer make money. The St. dev.: 11.36% 17.85% 21.91% 22.75% 21.23%
reason, they argued, was that the markets had become so
sophisticated over the last several years that whatever ineffi- Annualized 12 24 36 48 60
ciencies made the strategy profitable during the 1970s and months months months months months
1980s had been eliminated. Most recent: 9.97% 8.17% 6.40% 0.06% 0.57%
Average: 2.53% 2.30% 1.40% 0.73% 0.43%
STRATEGY SUMMARY Best: 44.61% 21.40% 13.95% 10.73% 8.40%
Profitability Trade statistics Worst: -20.11% -10.48% -9.43% -8.08% -5.51%
End. equity ($): 123,221 No. trades: 199 St. dev.: 11.36% 8.56% 6.83% 5.26% 3.93%
Total return (%): 23 Avg. trade ($): 117
Avg. annual ret. (%): 2.20 Avg. DIT: 90.1 LEGEND: Cumulative returns — Most recent: most recent return from start to
Profit factor: 1.03 Avg. win/loss ($): 2,028 (1,045) end of the respective periods • Average: the average of all cumulative returns
from start to end of the respective periods • Best: the best of all cumulative returns
Avg. tied cap (%): 40 Lrg. win/loss ($): 21,163 (2,832)
from start to end of the respective periods • Worst: the worst of all cumulative
Win. months (%): 52 Win. trades (%): 32.2 returns from start to end of the respective periods • St. dev.: the standard devia-
Drawdown TIM (%): 100 51.2 tion of all cumulative returns from start to end of the respective periods
Max. DD (%): 32.9 Tr./Mark./Year: 1.4 Annualized returns — The ending equity as a result of the cumulative returns,
raised by 1/n, where n is the respective period in number of years
Longest flat (m): 82.6 Tr./Month: 1.7
Send Active Trader your systems
LEGEND: End. equity ($) — equity at the end of test period • Total return If you have a trading system or idea you’d like tested, send it to
(%) — total percentage return over test period • Avg. annual ret. (%) — us at the Trading System Lab. We’ll test it on a portfolio of
average continuously compounded annual return • Profit factor — gross stocks or futures (for now, maximum 60 markets, using the last
profit/gross loss • Avg. tied cap (%) — average percent of total available cap- 2,500 trading days), using true portfolio analysis/optimization.
ital tied up in open positions • Win. months (%) — percentage profitable Most system-testing software only allows you to test one mar-
months over test period • Max. DD (%) — maximum drop in equity • ket at a time. Our system-testing technique lets all markets
Longest flat — longest period, in months, spent between two equity highs • share the same account and is based on the interaction within
No. trades — number of trades • Avg. trade ($) — amount won or lost by the portfolio as a whole.
the average trade • Avg. DIT— average days in trade • Avg. win/loss ($) Start by e-mailing system logic (in TradeStation’s
— average winning and losing trade, respectively • Lrg. win/loss ($) — EasyLanguage or in an Excel spreadsheet) and a short description
largest winning and losing trade, respectively • Win. trades (%) — percent to editorial@activetradermag.com, and we’ll get back to you.
winning trades • TIM (%) — amount of time there is at least one open posi- Note: Each system must have a clearly defined stop-loss level
tion for entire portfolio, and each market, respectively • Tr./Mark./Year —
and a suggested optimal amount to risk per trade.
trades per market per year • Tr./Month — trades per month for all markets

Disclaimer: The Trading System Lab is intended for educational purposes only to provide a perspective on different market concepts. It is not meant to recommend or
promote any trading system or approach. Traders are advised to do their own research and testing to determine the validity of a trading idea. Past performance does not
guarantee future results; historical testing may not reflect a system’s behavior in real-time trading.

ACTIVE TRADER • February 2003 • www.activetradermag.com 63


FUTURES
Trading System Lab
FIGURE 1 COMPARING THE STOPS
Notice how the modified exit rule approaches price at an accelerated
rate as time passes.
Experimenting with exits Nasdaq 100 index (ND), daily 900.00
880.00
Market: Futures. 860.00
840.00
System concept: Some traders believe exit signals are more
important than entry signals, while others believe success Sell 820.00
hinges on money management and diversification. The 800.00
truth is most likely somewhere in the middle and, as a 780.00
result, all components of a trading system must be ade- 760.00
quately developed and tested. 740.00
This system’s entry technique is simply a breakout above
720.00
a 55-day high. The more important part of the strategy is a
trailing stop technique that is designed to capture as much 700.00
of the trend as possible by tightening the stop relative to the 680.00
number of days the trade has been open. 660.00
After initiating a trade, a recent low point, such as the 640.00
lowest low of the past 55 days, is selected. To determine the Buy
620.00
stop level, the 20-day average true range (ATR) is multi-
600.00
plied by 10 percent, and then multiplied by the number of
days the trade has been open. This amount is then added to 580.00
the recent low used as the initial reference point. 560.00
For example, if a trade has been open for 12 days, we December 1997 January 1998 Feb. 1998
would multiply the 20-period ATR by 0.1 (10 percent), mul-
Source for all figures: Wealth-Lab Inc. (www.wealth-lab.com)
tiply this result by 12 (days) and, finally, add the result to
the lowest low of the past 55 days. The next day, the 20-day ATR Rules:
would be multiplied by 0.1, multiplied by 13, added to the lowest 1. Enter long on the next bar’s open if today’s high is
low of the past 55 days, and so on. greater than the highest high of the last 55 days.
We will compare the results of this stop with using the lowest 2. Exit on the next day’s open if the lowest low is less than
low of the past 55 days. Figure 1 shows the difference between the the lowest low of the last 55 days plus 10 percent of the
two stops. The magenta line represents the 55-day low stop and the actual average true range for each day in the market.
blue line is the modified exit strategy. The blue line tracks the price
action more closely. This is a long-only system.

FIGURE 2 EQUITY CURVE (MODIFIED EXIT RULE) FIGURE 3 EQUITY CURVE (SIMPLE EXIT RULE)
The modified exit rule produced modest profits The simple 55-day low exit technique actually
over the test period. outperformed the modified stop.
650,000 1,000,000
600,000 950,000
900,000
550,000 850,000
800,000
500,000
750,000
450,000 700,000
650,000
400,000
600,000
Account balance ($)

550,000
Account balance ($)

350,000
500,000
300,000 450,000
250,000 400,000
350,000
200,000 300,000
150,000 250,000
200,000
100,000 150,000
100,000
50,000
50,000
0 0
3/22/94 5/1/95 6/3/96 7/3/97 8/3/98 9/1/99 11/1/00 1/2/02 2/3/03
3/22/94 5/1/95 6/3/96 7/3/97 8/4/98 9/3/99 11/1/00 1/2/02 2/3/03

Equity Cash Linear Reg. Long Short Equity Cash Linear Reg. Long Short

64 www.activetradermag.com • June 2004 • ACTIVE TRADER


FIGURE 4 DRAWDOWN
The modified exit rule succeeded in reducing the system’s risk (28 percent
Money management: Risk a maximum 3 per- compared to 36 percent).
cent of total account equity per trade (“stop- 0.00%
based risk %”). Setting the “maximum risk -2.00%
percent” to 3 means at any given time we -4.00%
should not lose more than 3 percent of the total -6.00%
account equity. Of course, overnight gaps and -8.00%

Drawdown
limit down moves could lead to higher losses. -10.00%
-12.00%
Starting equity: $250,000 (nominal). Deduct -14.00%
$20 slippage/commission per round-turn
-16.00%
trade.
-18.00%

Test period: March 1994 to August 2003. -20.00%


-22.00%
Test data: The system was tested on the Active -24.00%
Trader Standard Futures Portfolio, which con- -26.00%
tains the following 19 futures: DAX30 (AX), -28.00%
corn (C), crude oil (CL), German bund (DT), 3/22/94 5/1/95 6/3/96 7/1/97 8/3/98 9/1/99 11/1/00 1/2/02 2/3/03
euro dollar (ED), euro forex (FX), gold (GC),
copper (HG), Japanese yen (JY), coffee (KC),
live cattle (LC), lean hogs (LH), Nasdaq 100 (ND), natural gas uses the simple 55-day low exit rule. This version had fewer trades
(NG), soybeans (S), sugar (SB), silver (SI), S&P 500 (SP) and T- (208) than the modified system because the stop maintains a con-
Notes 10 year (TA). The test used ratio adjusted data from Pinnacle stant distance from price action regardless of volatility changes
Data Corp. and how long the trade has been open. The average profit per trade
increased to $3,073 and the average holding time increased from 40
System results: Figure 2 shows the equity curve when risking 3 days to 95 days. The simple system’s net profit was much higher
percent of the total portfolio equity per trade. The system returned ($639,204 compared to $338,282), although its drawdown was also
a total profit of 135.31 percent over approximately 10 years and markedly higher (36.65 percent compared to 28.7 percent).
9.54 percent annually, with the worst year being a loss of 8.3 per-
cent in 1999. The single largest annual drawdown of 19.99 percent Bottom line: The trailing exit strategy enabled the system to
occurred in 1996. The system produced 351 trades with an average reduce drawdown, but it also reduced profits. Comparison to the
profit of $963.77 per trade. simple stop approach suggests the modified version exited trades
Figure 3 shows the equity curve of the comparison system that too quickly and did not give the system enough time to ride the
trend.
STRATEGY SUMMARY However, the concept behind this exit idea is worthy of exper-
Profitability Trade statistics imentation and could prove to be a more effective exit strategy
Net profit ($): 338,282.00 No. trades: 351 when combined with other trading methods.
Net profit (%): 135.31 Win/loss (%): 42.17
Exposure (%): 33.88 Avg. gain/loss (%): 0.65 — Volker Knapp of Wealth Lab
Profit factor: 1.32 Avg. hold time: 40.03
PERIODIC RETURNS
Payoff ratio: 1.70 Avg. profit (winners) %: 7.97
Recovery factor: 2.56 Avg. hold time (winners): 58.09 Avg. Sharpe Best Worst Percentage Max. Max.
Drawdown Avg. loss (losers) %: -4.69 return ratio return return profitable consec. consec.
periods profitable unprofitable
Max. DD (%): -28.70 Avg. hold time (losers): 26.87
Weekly 0.21% 0.59 11.95% -9.16% 55.19% 9 8
Longest flat days: 760 Max. consec. win/loss: 6/9
Monthly 0.88% 0.59 16.74% -9.95% 52.63% 6 4
LEGEND: Net profit — Profit at end of test period, less commission • Quarterly 2.60% 0.57 22.78% -10.91% 56.41% 4 3
Exposure — The area of the equity curve exposed to long or short positions, Annually 9.86% 0.63 40.57% -8.30% 70.00% 7 1
as opposed to cash • Profit factor — Gross profit divided by gross loss •
Payoff ratio — Average profit of winning trades divided by average loss of los- LEGEND: Avg. return — The average percentage for the period • Sharpe
ing trades • Recovery factor — Net profit divided by max. drawdown • ratio — Average return divided by standard deviation of returns (annual-
Max. DD (%) — Largest percentage decline in equity • Longest flat days — ized) • Best return — Best return for the period • Worst return — Worst
Longest period, in days, the system is between two equity highs • No. trades return for the period • Percentage profitable periods — The percentage of
— Number of trades generated by the system • Win/Loss (%) — The per- periods that were profitable • Max. consec. profitable — The largest num-
centage of trades that were profitable • Avg. gain — The average profit for all ber of consecutive profitable periods • Max. consec. unprofitable — The
trades • Avg. hold time — The average holding period for all trades • Avg. largest number of consecutive unprofitable periods
gain (winners) — The average profit for winning trades • Avg. hold time
Trading System Lab strategies are tested on a portfolio basis (unless
(winners) — The average holding time for winning trades • Avg. loss (los-
otherwise noted) using Wealth-Lab Inc.’s testing platform.
ers) — The average loss for losing trades • Avg. hold time (losers) — The
If you have a system you’d like to see tested, please send the trad-
average holding time for losing trades • Max. consec. win/loss — The max-
ing and money-management rules to editorial@activetradermag.com.
imum number of consecutive winning and losing trades

Disclaimer: The Trading System Lab is intended for educational purposes only to provide a perspective on different market concepts. It is not meant to recommend or
promote any trading system or approach. Traders are advised to do their own research and testing to determine the validity of a trading idea. Past performance does not
guarantee future results; historical testing may not reflect a system’s behavior in real-time trading.

ACTIVE TRADER • June 2004 • www.activetradermag.com 65


FUTURES
Trading System Lab
Monthly breakout
Market: Futures. Optimization parameters: The x and y parameters will be
optimized from one to eight months.
System concept: Some traders believe designing a robust trading
system requires dividing your data into two sets. The first half Test data: Daily continuous T-Bond futures (US). No com-
(referred to as “in-sample” data) is used to develop trading rules missions or slippage were deducted.
and optimize their parameters; the second (“out-of-sample” data)
is used to simulate trading the system and see if the results are Test periods: Initial “in-sample” period: Jan. 1, 1985, through
both favorable and consistent with the initial test. Dec. 31, 1995. Second “out-of-sample” period: Jan. 1, 1996, to
If the system performs well on the second data set (which rep- Dec. 31, 2002.

FIGURE 1 OPTIMIZATION RESULTS FROM IN-SAMPLE TEST


The results for different parameter combinations are sorted by “recovery factor,” which is the net profit divided by the maximum
drawdown.

resents new, “unseen” price action), the system is considered to Test results — In-sample: All parameter combinations were
have a better chance of working in real trading. The process is profitable. Figure 1 (above) shows the top combinations sorted
referred to as “walk-forward testing” because the system can be by “Recovery Factor,” which is the absolute value of the sys-
progressively applied to new data to see if it continues to perform. tem’s net profit divided by its maximum drawdown (see the
We will explore that concept here. To illustrate the principles stock Trading System Lab on p. 56 for the significance of this
in a straightforward fashion, we test a basic, long-only monthly statistic).
breakout system on a single market, the T-Bond.
Even though the monthly highs and lows define the FIGURE 2 PROFIT CURVE
entry points, we will use end-of-day data to take The optimized monthly breakout system slightly underperformed
opening gaps into consideration. buy-and-hold, but it had a much lower risk level.
Whenever performing out-of-sample testing,
40,000
expect the performance to be worse than the in-sam-
35,000
ple tests. This is not a reflection of the quality of the
30,000
Account balance ($)

system. It is simply because the system was devel-


25,000
oped and optimized on a different data set.
20,000
15,000
Rules:
10,000
1. Entry: Buy at the highest high of the last x
months. 5,000
2. Exit: Sell at the lowest low of the last y months. 0
1/2/85 12/2/85 1/2/87 1/4/88 1/3/89 1/2/90 1/2/91 1/2/92 1/4/93 1/3/94 1/3/95
Total profit Buy & hold Linear reg
Risk control: The system does not use a fixed dollar,
point or percentage stop. Instead, a reversal below the Source for all figures: Wealth-Lab Inc. (www.wealth-lab.com)
y-month low is used to indicate the market is no
longer in an uptrend, at which point all positions are liquidated, For the 10 years covered in the test, the best parameters
win or lose. would be to buy at the highest high of the past month and
sell at the lowest low of the past two months. As expected,
Money management: Each position will consist of one T-bond the shorter the breakout length, the higher the number of
contract. trades. However, the optimal settings did not produce many
trades (19) over the test period.

66 www.activetradermag.com • March 2004 • ACTIVE TRADER


FIGURE 3 OUT-OF-SAMPLE TEST RESULTS
The system behaved much differently on the out-of-sample data. The best parameter combination from the in-sample test was the third
worst in the out-of-sample test.

FIGURE 4 PROFIT CURVE


The optimized parameters from the in-sample test performed Bottom line: The testing process illustrated here
much worse on the out-of-sample data. shows how the performance of even fundamentally
sound trading systems can vary over time. The best
20,000 parameter set of the first 10 years became the third
15,000 worst in the following seven years. However, all the
Account balance ($)

10,000 parameter combinations were profitable in both peri-


ods, suggesting the basic trading approach is sound.
5,000
The best-performing parameters in one period will
0 almost never be the best in future data. Because of this,
-5,000 it is more important to look for parameter stability. A
1/2/96 7/29/96 3/31/97 12/1/97 7/28/98 3/30/99 12/1/99 7/27/00 3/28/01 12/3/01 8/1/02 broad range of parameter values should be profitable
and deliver consistent results.
Total profit Buy & hold Linear reg Optimizing is a very useful tool, but it should be
used to confirm parameter stability rather than to
The equity curve (dark green) in Figure 2 (opposite page) for try to find the parameter combination with the highest net
the most part follows the underlying market curve (blue line). profit.
Even though the buy-and-hold profit ($44,006) over the 10-year
in-sample period was higher than the system’s profit ($40,012), — Dion Kurczek and Volker Knapp of Wealth-Lab Inc.
the buy-and-hold drawdown was more than 50 percent higher
($10,886) than the system’s drawdown ($6,428). FIGURE 5 SAMPLE TRADE
The system performed best in the out-of-sample test when it bought
Test results — Out-of-sample data: Testing the parameters on the above the one-month high and sold below the eight-month low.
out-of-sample data set produced significantly different results.
106.00
The best parameter combination from the in-sample test (buy T-bonds (US), daily 105.50
above the one-month high and sell below the two-month low) 105.00
became the third worst combination in the out-of-sample test. 104.50
Figure 3 (top) sorts the 11 worst parameter combinations by 104.00
recovery factor. Two combinations from the former top 11 are 103.50
Sell
now in the bottom eleven. On the other hand, all the parame- 103.00
102.50
ter combinations have remained profitable, which indicates
102.00
the system rules have some merit. 101.50
Figure 4 (above) shows the equity curve for the optimized 101.00
parameters from the in-sample data period tested on the out- 100.50
of-sample period. Even though the performance was positive, 100.00
it was quite volatile compared to the smooth ride of the opti- 99.50
99.00
mal parameter combination (buying above the one-month
98.50
high and selling below the eight-month low) for this test peri- 98.00
od, a sample trade of which is shown in Figure 5 (right). Buy
97.50
97.00
Trading System Lab strategies are tested on a portfolio basis (unless 96.50
otherwise noted) using Wealth-Lab Inc.’s testing platform. 96.00
If you have a system you’d like to see tested, please send the trad- Volume
95.50
ing and money-management rules to editorial@activetradermag.com. August 2001 September 2001 October 2001 November 2001

Disclaimer: The Trading System Lab is intended for educational purposes only to provide a perspective on different market concepts. It is not meant to recommend or
promote any trading system or approach. Traders are advised to do their own research and testing to determine the validity of a trading idea. Past performance does not
guarantee future results; historical testing may not reflect a system’s behavior in real-time trading.

ACTIVE TRADER • March 2004 • www.activetradermag.com 67


60-minute breakout system
FIGURE 1 EQUITY CURVE
Markets: Stocks.
We tested the portfolio of 18 stocks on 30-minute bars (more than 3,500
total trades). Each signal used five percent of the portfolio equity value.
System concept: This intraday system takes a trade
Long trades were slightly profitable.
when price breaks out of the trading range established
160,000
in the first hour of the session. Early in the morning the
150,000
market often is trying to establish its direction, and a
140,000
move above or below the 60-minute range might sig- 130,000
nal a trend in that direction. Also, a breakout of the 120,000
early trading range is sometimes caused by a specific 110,000
news item that will cause the trend to continue. 100,000
Because the highest volatility of the day often occurs in 90,000
the first trading hour, there are no trades in the first 60 80,000
minutes. 70,000
After the first 60 minutes have ended, if the closing 60,000
price of the current 30-minute bar is above the high of 50,000
the range, we will go long on the next open. A short 40,000
position is established if the closing price is below the 30,000
20,000
range. A signal in the opposite direction is used to exit
10,000
the current position. All open positions are exited at
0
the close of the day. There will be only one trade per 10/10/02 11/25/02 1/14/03 2/28/03 4/14/03 5/29/03 7/14/03 8/26/03 10/8/03
day. Equity Cash Linear reg Long Short Buy & hold

Entry rules:
Long trades: Buy if the closing price of the third 30-minute Test period: October 2002 through October 2003.
bar is above the high of the first 60 minutes of the day.
Initial test results: The system’s performance was disappoint-
Short trades: Sell short if the closing price of the third 30- ing — an overall loss of -3.25 percent. With 3,546 trades in the
minute bar is below the low of the first 60 minutes of the day. test period, this is one of the more active systems tested here.
However, even setting the commission at one cent per share
Exit: Exit all positions on signals in the opposite direction or (advisable for a system that generates this many trades) still
at the end of the day. resulted in $13,576 in commission charges. Figure 1 shows
long trades were slightly profitable, generating a 4.22-percent
Money management: Each trade is sized at five percent of the profit during this period.
current account equity. This will allow all trades the
system is generating to be executed. Increasing the
percentage would require dropping trades that FIGURE 2 ADDING A FILTER
exceeded the available cash limit. The bars show the average per-trade profit that would have been
captured by adding the CMO as a trade filter.
Starting equity: $100,000 (nominal). Deduct $0.01 per 0.80%
share slippage and commissions.
0.70%

Test data: The system was tested on 30 minute bars of 0.60%


the Active Trader Standard Stock Portfolio, which con-
0.50%
tains the following 18 stocks: Apple Computer
(AAPL), Boeing (BA), Citibank (C), Caterpillar (CAT), 0.40%
Cisco (CSCO), Disney (DIS), General Motors (GM),
0.30%
Hewlett Packard (HPQ), International Business
Machines (IBM), Intel (INTC), International Paper 0.20%
(IP), JPMorgan Chase (JPM), Coke (KO), Microsoft -80.00 -60.00 -40.00 -20.00 0.00 20.00 40.00 60.00 80.00
(MSFT), Sears (S), Starbucks (SBUX), AT&T (T) and CMOlevel
Wal-Mart (WMT). Data from www.qcharts.com.

68 www.activetradermag.com • January 2004 • ACTIVE TRADER


FIGURE 3 SAMPLE TRADE
Adding a filter: To see if we could reduce the number of trades The green bars are the first trading hours of the days shown here.
and improve performance, we added a filter — a 14-bar Subsequent black bars have closing prices within the first-hour
Chande Momentum Oscillator (CMO) to the basic system (see range; red bars indicate the closing price is below the range and
the December 2003 Trading System Lab, p. 48, for more infor- blue bars show bars with closing prices above the range.
mation on this indicator). The filter consisted of taking trades
Cisco Systems (CSCO), 30-minute 21.50
only when the CMO reading for the previous day’s bar was 21.40
above 60, indicating bullish momentum. The filter reduced the 21.30
total number of trades and eliminated short trades. Figure 2 Short 240 21.20
@20.87 21.10
(opposite page) shows waiting for a CMO reading above 60
21.00
before taking trades would have increased profitability. 20.90
Testing our portfolio with the additional filter turned the los- 20.80
ing system into a winner — although not a spectacular one. 20.70
The number of trades decreased to 251 from 3,546, the average 20.60
20.50
profit was 0.12 percent and total profit was 1.48 percent. 20.40
20.30
Bottom line: Although the strategy concept sounds reasonable, 20.20
it did not produce satisfactory results. There is plenty of room 20.10
Cover 240 Buy 246 20.00
for further experimentation, though — such as trying other fil- @20.34 @20.34 19.90
ters, including those that would differentiate between long and
9/24/03 9/25/03
short trades. Another alternative is to use shorter bars (e.g., five
minutes) to get into positions faster.
Although some individual stocks produced very good
equity curves, the total portfolio itself did not. This illustrates
how testing a strategy on a single instrument or very limited
STRATEGY SUMMARY portfolio can lead to the wrong conclusions about a system’s
Profitability Trade statistics value.
Net profit ($): -3,247.41 No. trades: 3,546
Net profit (%): -3.25 Win/loss (%): 49.55 — Volker Knapp of Wealth-Lab Inc.
Exposure (%): -3.26 Avg. gain/loss (%): -0.02
Profit factor: 0.96 Avg. hold time: 6.46
Payoff ratio: 0.98 Avg. profit (winners) (%): 0.85 PERIODIC RETURNS
Recovery factor: 0.42 Avg. hold time (winners): 6.71
Avg. Sharpe Best Worst Percentage Max. Max.
Drawdown Avg. loss (losers) (%): -0.87 return ratio return return profitable consec. consec.
periods profitable unprofitable
Max. DD (%): -7.45 Avg. hold time (losers): 6.22 Weekly -0.06% -0.56 1.70% -2.11% 47.17% 4 6
Longest flat days: 3,078 Max. consec. win/loss: 18/23
Monthly -0.24% -0.60 2.48% -2.57% 38.46% 2 5
LEGEND: Net profit — Profit at end of test period, less commission • Exposure Quarterly -0.63% -0.51 1.99% -4.47% 60.00% 1 1
— The area of the equity curve exposed to long or short positions, as opposed to cash
• Profit factor — Gross profit divided by gross loss • Payoff ratio — Average
LEGEND: Avg. return — The average percentage for the period • Sharpe ratio —
profit of winning trades divided by average loss of losing trades • Recovery factor
Average return divided by standard deviation of returns (annualized) • Best return
— Net profit divided by max. drawdown • Max. DD (%) — Largest percentage
— Best return for the period • Worst return — Worst return for the period •
decline in equity • Longest flat days — Longest period, in days, the system is
Percentage profitable periods — The percentage of periods that were profitable •
between two equity highs • No. trades — Number of trades generated by the sys -
Max. consec. profitable — The largest number of consecutive profitable periods •
tem • Win/Loss (%) — the percentage of trades that were profitable • Avg. prof-
Max. consec. unprofitable — The largest number of consecutive unprofitable peri -
it — The average profit for all trades • Avg. hold time — The average holding peri -
ods
od for all trades • Avg. profit (winners) — The average profit for winning trades
• Avg. hold time (winners) — The average holding time for winning trades • Trading System Lab strategies are tested on a portfolio basis (unless
Avg. loss (losers) — The average loss for losing trades • Avg. hold time (losers) otherwise noted) using Wealth-Lab Inc.’s testing platform.
— The average holding time for losing trades • Max. consec. win/loss — The If you have a system you’d like to see tested, please send the
maximum number of consecutive winning and losing trades trading and money-management rules to editorial@activetradermag.com.

Disclaimer: The Trading System Lab is intended for educational purposes only to provide a perspective on different market concepts. It is not meant to recommend or
promote any trading system or approach. Traders are advised to do their own research and testing to determine the validity of a trading idea. Past performance does not
guarantee future results; historical testing may not reflect a system’s behavior in real-time trading.

ACTIVE TRADER • January 2004 • www.activetradermag.com 69