MTA JOURNAL
Spring-Summer 2001
Issue 55
A Publication of
MARKET TECHNICIANS ASSOCIATION, INC.
One World Trade Center
Suite 4447
212/912-0995
Fax: 212/912-1064
e-mail: shelley@mta.com
Incorporated 1973
www.mta.org
ISSUE 55
EDITOR'S COMMENTARY
35
41
WYCKOFF TESTS: NINE CLASSIC TESTS FOR ACCUMULATION; NINE NEW TESTS FOR RE-ACCUMULATION
29
21
Blethyn Hulton
51
MTA JOURNAL
Spring-Summer 2001
ISSUE 55
EDITOR
Henry O. Pruden, Ph.D.
Golden Gate University
San Francisco, California
ASSOCIATE EDITOR
David L. Upshaw, CFA, CMT
Lake Quivira, Kansas
MANUSCRIPT REVIEWERS
Connie Brown, CMT
Aerodynamic Investments Inc.
Pawley's Island, South Carolina
John A. Carder, CMT
Topline Investment Graphics
Boulder, Colorado
Ann F. Cody, CFA
Hilliard Lyons
Louisville, Kentucky
Cynthia Kase
Kase and Company
Albuquerque, NM
PRODUCTION COORDINATOR
Barbara I. Gomperts
Financial & Investment Graphic Design
Marblehead, Massachusetts
PUBLISHER
Market Technicians Association, Inc.
One World Trade Center, Suite 4447
New York, New York 10048
MTA JOURNAL
Spring-Summer 2001
MTA JOURNAL
Spring-Summer 2001
AFFILIATE
Affiliate status is available to individuals who are interested in technical analysis, but who do not fully meet
the requirements for Member, as stated above; or who currently do not know three MTA members for sponsorship. Privileges are noted below.
DUES
Dues for Members and Affiliates are $200 per year and are payable when joining the MTA and thereafter
upon receipt of annual dues notice mailed on July 1. College students may join at a reduced rate of $50 with the
endorsement of a professor.
APPLICATION FEES
Applicants for Member will be charged a onetime, nonrefundable application fee of $25; no fee for Affiliates.
Members
Affiliates
MTA JOURNAL
Spring-Summer 2001
Management Committee
Director: President
Accreditation
Journal
Body of Knowledge
Membership
Richard A. Dickson
Hilliard Lyons
502/588-4122
Fax: 502/588-9132
E-mail: rdickson@hilliard.com
Director: Secretary
Bruno DiGiorgi
Lowry's Reports Inc.
561/842-3514
Fax: 561/842-1523
E-mail: bdigiorgi@lowrysreports.com
Director: Treasurer
Andrew Bekoff
Van Der Moolen LLC
212/495-0558
Fax: 212/809-9143
E-mail: abekoff@bloomberg.net
Director: Past President
Library
Larry Katz
Market Summary & Forecast
805/370-1919
Fax 805/777-0044
E-mail: lk1618@mta.org
Newsletter
Richard A. Dickson
Hilliard Lyons
502/588-4122
Fax: 502/588-9132
E-mail: rdickson@hilliard.com
Michael N. Kahn
Bridge Information Systems
212/372-7541
E-mail: mkahn@bridge.com
Education
Rick Bensignor
Morgan Stanley Dean Witter
212/761-6148
Fax: 212/761-0471
E-mail: rick.bensignor@msdw.com
Mike Epstein
NDB Capital Markets Corp.
617/753-9910
Fax: 617/753-9914
E-mail: mepstein@ndbcap.com
Internship Committee
MTA JOURNAL
Spring-Summer 2001
Placement
Programs (NY)
Bernard Prebor
MCM MoneyWatch
212/908-4323
Fax: 212/908-4331
E-mail: bprebor@mcmwatch.comt
Regions
Nina G. Cooper
Pendragon Research, Inc.
815/244-4451
Fax: 815/244-4452
E-mail: ngcooper@grics.net
EDITOR'S COMMENTARY
MULTI-METHOD RESEARCH
Henry O. (Hank) Pruden, Ph.D., Editor
As technical analysts strive to raise their body of knowledge and discipline to higher
standards, the canons of the scientific method and empirical testing will become more
and more commonplace. In earlier Editorials, the case was made for the formulation
and testing of Theoretical Models. In this issue there are examples of the two polar
approaches to doing empirical research in the social sciences, hence in behavioral finance and technical analysis. These two poles can be characterized as survey research
findings across a large number of instances to establish statistical validity and reliability
on the one extreme, and the individual case study on the other pole. The individual
case study illustrates numerous variables and contingencies operating together in a
real-life situation.
Most of our MTA Journal articles rely upon the survey type of technique, which is
appropriate and excellent. But there remains an art form to the execution of trades
and a complexity of interacting indicators and decisions that are better revealed through
the story of a case study. In this issue of the MTA Journal the article on Wyckoff
Rules is an exposition based upon a case study. Other technical analysts are encouraged to consider using the single-case approach to document their studies, perhaps
relating case studies based upon their own experiences.
MTA JOURNAL
Spring-Summer 2001
EDITOR'S COMMENTARY
ANOTHER PERSPECTIVE
I have a great respect for orthodoxy; not for those orthodoxies which prevail in particular
schools or nations, and which vary from age to age, but for a certain shrewd orthodoxy which the
sentiment and practice of laymen maintain everywhere. I think that common sense, in a rough
dogged way, is technically sounder than the special schools of philosophy, each of which squints
and overlooks half the facts and half the difficulties in its eagerness to find in some detail the key
to the whole. I am animated by distrust of all high guesses, and by sympathy with the old prejudices and workaday opinions of mankind: they are ill expressed, but they are well grounded.
My philosophy is justified, and has been justified in all ages and countries by the facts before
every man's eyes; and no great wit is requisite to discover it, only (what is rarer than wit) candor
and courage. Learning does not liberate men from superstition when their souls are cowed or
perplexed; and, without special learning, clear eyes and honest reflection can discern the hang
of the world, and distinguish the edge of truth from the might of imagination.
George Santayana, Preface to a New Philosophy
MTA JOURNAL
Spring-Summer 2001
INTRODUCTION
The conventional wisdom of technical analysis maintains that Volume (Vol) and Open Interest (OI) studies are breadth indicators.
Each measures the continued health, or potential weakening, of a
given trend. We expect increases in Vol and OI as prices move in the
direction of the primary trend, while we tolerate decreases in each as
prices move contrary to that trend. This application of Vol and OI
data is clear in the technical analysis literature.
My work suggests that more valuable and timely applications of
OI exist in the exchange-traded futures markets, specifically the US
Treasury Bond Futures market. In addition to providing broad confirmation of trend, or signaling pending divergence to the trend, OI
figures provide vital insight into sentiment levels within particular
markets. As such, OI provides information which is often contrary in
nature and more appropriately utilized as a shorter-term sentiment
gauge rather than as a longer-term breadth indicator. This paper
will show how OI data analysis can be extended into the area of market sentiment and contrary opinion, through a greater dissection of
the Commodity Futures Trading Commissions Commitment of Traders Reports. It will demonstrate ways to decompose changes in OI
that better reflect the true nature of positions in the UST Bond Futures market, and offer methodologies that can be used to forecast
price action on a shorter-term basis.
Volume
Open
Interest
Market
Interpretations
RISING
Increasing
Increasing
STRONG
Decreasing
Decreasing
WEAK
Increasing
Increasing
STRONG
Decreasing
Decreasing
WEAK
FALLING
The inherent delay between Breadth divergence and price reversals makes it difficult to translate breadth analyses into short-run trading strategies. Some of the more recent work on OI has attempted to
improve upon the Breadth framework in this regard. Hadady (1989)
developed the Price Open Interest (POI) Indicator to address this
issue. Hadady characterized percentage changes in Price and OI along
a Bullish-to-Bearish continuum. He reasoned that during those periods in which OI diverged from price action, bullish and bearish signals could be derived from the relative percentage changes of Price
and OI. Though intellectually appealing, application and interpretation are difficult for a number of reasons.
Figure 3 Interpretation of Price /OI Principles
OI analysis can be conducted under two distinct, but complementary, frameworks: a Breadth framework and a Money Flow framework. Each framework interprets changes in Total OI as an indication of whether money is flowing into or out of a given market. Rising OI represents money flowing into a market, while falling OI represents money flowing out. From this point of common interpretation, the two frameworks quickly diverge.
Figure 1 Formation of OI
Sale
New Long
New Short
Increases
New Long
Old Long
Unchanged
Old Short
New Short
Unchanged
Old Short
Old Long
Decreases
III.
IV.
OI Effect
MTA JOURNAL
I.
II.
Painter (1995) addressed the same issue with his On-Balance Open
Interest Indicator (OBOII), a variant of Joseph Granvilles On-Balance Volume (OBV). Painter used the major tenets and principles of
OI to create a time series which provided signals of confirmation and
divergence on a cumulative basis. His methodology incorporated
the use of dummy variables (not derived from actual market data)
which were generated by four different market states. He used these
variables to construct a cumulative index that identified changes to
the intermediate trend. Though it provides an ongoing, cumulative
measure of breadth, the inherent weaknesses of a breadth framework
remain.
A Breadth Framework
A Breadth framework does not differentiate between the activity
of different participants. All new money (rising OI) supports the
primary trend, regardless of who is establishing new positions and on
which side of the market they are doing so. So long as new positions
are established, there is fuel available to feed the trend. This underscores the longer term nature of a Breadth framework, where the
goal is to identify signals of confirmation of or divergence from to
the longer-term trend.
Purchase
Underlying Reasoning
Spring-Summer 2001
Exhibit 2
MTA JOURNAL
Spring-Summer 2001
10
The utility of the CoT Report is derived from two measures: Total
Absolute Net Positions and the positioning of Comml versus Specs and STs.
[I refer to the combined positions of Specs and ST as FADE positions.] Any decomposition must capture and measure both to be
effective. Total Absolute Net Positions (TAbs Net Pos) gauge the
magnitude of conflicting opinion in the market by measuring how
far from Neutral the net positions of adversarial participants have
moved, while the relative positioning of Comml vs. Specs/ST depicts
the polarity (long v. short) of the smart money.
Although there is no identity property that states the Spec and ST
categories must align, they do so more often than not, especially as
TAbs Net Pos increase. For example, the Spec and ST components
are aligned similarly in 304, or 63%, of the 482 weekly observations
in the CoT Report for bonds between January 1986 and November
1998. The observations that do not align occur in periods of relatively low TAbs Net Positions. Specifically, of the 178 (37%) observations in which the Net Pos of Specs and NRPT are not aligned, 153,
or 86%, occur when TAb Net Positions are less than 15% of Spread
Adjusted OI. As can be seen in Exhibit 3, this is a relatively low reading for TAbs Net Pos in the CoT Report for bonds. Simply put, when
the magnitude of disagreement is small, the relative positioning of
the players matters less.
With this in mind, a more appropriate CoT decomposition for
bonds is provided in Exhibit 4. The first graph shows TAbs Net Pos
between January 1986 and November 1998. The second and third
graphs show Net Comml Pos and Net FADE Pos, respectively. Although the sum of the absolute values of the second and third graph
will not always equal the first, the latter two graphs are now perfect
offsets. This confirms that the strong- and weak-handed components
are correctly accounted for, even if they fail to perfectly align on opposite sides of the market in every observation.
In order to remove the trends caused by dramatic increases in OI
over the last few years (Figure 4), each category is restated as a percentage of Spread Adjusted OI (defined in Appendix A) in Exhibit 5.
This adds a degree of relativity to the time series for purposes of historical comparison. These OI-adjusted time series are defined in Figure 5. Further, Exhibit 5 clarifies the relationship between TAbs Net
Pos and CoT%C & CoT%FADE. As TAbs Net Pos increase, conflicting opinions/positions in the market are on the rise. As OI is buoyed
by this increase in TAbs Net Positions, the market position of the
strong versus weak hand of the market becomes more polar or extreme. As the exhibit shows, CoT%C and CoT%FADE can move to
either extreme as TAbs Pos are forced higher.
CoT%ST:
CoT%C:
CoT%S:
CoT%Fade:
Exhibit 4
Exhibit 3
MTA JOURNAL
Spring-Summer 2001
11
Exhibit 6
MTA JOURNAL
Spring-Summer 2001
12
Exhibits 7
Exhibits 7A
Box A
See Exhibits 7 and 7A. In Jan 98, bonds traded as high as 123.17 on a closing basis. Spread Adj OI
increased by approximately 32% between early Sep and mid-Jan as bonds were making their highs. TAbs
Net Pos rose from a neutral 1.0% to 31.6% at the high. Coincident to the increase in TAbs Net Pos,
CoT%C moved from .3% to -15.8%, a bearish extreme. This was a clear signal that bonds were spurred
higher on an increase in speculative activity. The increasing conflict between the strong and weak factions
of the market was underscored by a dramatic increase in TAbs Net Pos, and CoT%C maintained the
unwinding of conflicting opinion would lead to lower prices. By the time this conflict was unwound in
early Mar, and these indicators moved back to Neutral levels, bonds had fallen 4-3/4 points.
A similar scenario developed in Jan 96 as bonds traded to the year low yield of 5.92%. Bonds rallied 8
5/8 points between the end of Sep and early Jan 96 on a 28% increase in OI. As OI increased over that
period, TAbs Net Positions rose from 6.4% to 21.2%, suggesting increasing conflict and polarity in market
sentiment and positions. CoT%Cs move from 3.1% to -8.2% suggested this would be resolved through
lower Bond prices. Over the ensuing 4 months, Bond prices fell over 14 points as these indicators moved
to Neutral levels.
Price action in May 95 illustrates how similar increases in TAbs Net Positions led to dramatic rallies.
Bonds were fighting psychological resistance at 5.25% in May 95 following the 1994 Bear Market in bonds.
After rallying over 3-1/2 points from the beginning of the year to early Feb, bonds rallied an additional 31/8 points by early May 95. OI increased by 7.5% over the first part of the rally, and fell by approximately
2.5% over the latter part. As bonds rallied nearly 6-3/4 points between Jan and May95, TAbs Net Positions
increased from 13.9% to 26%. CoT%C increased from 6.9% in early Jan to 13.0% in early May following
only a slight moderation in early Feb. In this instance, CoT%C suggested the increased conflict in positioning would be resolved via higher prices since the strong hand of the market had continued to accumulate positions. The rally that ensued saw bonds rally by more than 9-1/2 points in little over four weeks.
MTA JOURNAL
Spring-Summer 2001
13
Exhibit 8
Exhibit 8A
BOX B
Jul 97 and Oct 98 are two times in which TAbs Net Positions did not mirror the
change in Spread Adjusted OI. [See Exhibit 8 and 8A]. In Jul 97 bonds were completing a rally begun in mid-Apr97. Spread Adj OI increased by 21% as bond prices climbed
nearly 8-1/2 points. Coincident to the increase in OI, TAbs Net Pos increased from
9.0% to 30.6% at the highs. As bond prices fell by 2-3/4 points in Aug, OI increased by
4%. However, as OI rose, TAbs Net Pos moved from 30.6% to 1.0%, in a dramatic
moderation in sentiment. As such, the market was unencumbered by speculative excess as prices began to climb in early September. This set the stage for further gains.
By the time TAbs Net Pos revisited the 30.8% area in mid-Dec, bonds were seven points
higher.
Oct98 represents a recent example of OI and TAbs Net Positions moving in opposite directions. In this instance, a dramatic decrease in Spread Adjusted OI was accompanied by an increase in TAbs Net Pos. During the liquidity crisis of Jul 98-Oct 98
bonds rallied smartly as a flight to quality put a bid under the Treasury market. Bond
prices rallied 10 points between the end of Jul and early Oct. Spread Adjusted OI,
which had already peaked in mid-Jun, fell an additional 15% between the end of Jul
and the high prices of early Oct. Over the same period, TAbs Net Pos rose to 36.4%
from 20.8%. Therefore, as money was leaving the market (OI falling), prices were
being driven higher by speculative forces (i.e. TAbs Net Pos rising; CoT%C falling).
The divergence between prices (up 10 pts) and OI (down 25%) quickly became acute
when TAbs Net Positions reached the 36.4% level. Prices reversed quickly thereafter.
MTA JOURNAL
Spring-Summer 2001
14
BIBLIOGRAPHY
Third, the relationship between TAbs Net Pos and Total OI outstanding can be used to determine whether significant corrections to
the trend are necessary and should be anticipated. TAbs Net Pos
typically increases as Total OI increases. This dynamic provides the
early signs that bad money is beginning to follow good money into
the market over the short-to-intermediate time frame. However, this
need not be the case. Increasing OI amidst flat or decreasing TAbs
Net Pos would suggest a price move that is sustainable over both the
short-term and the long-term. Increasing OI amidst increasing TAbs
Net Pos suggests the market may be suffering the ills of speculative
excess over the short-term. Whereas the former scenario warrants
relatively small corrections in price action, the latter warrants more
intense corrections. [Box B provides 2 examples of application.]
CoT%C Studies
In addition to contemporaneous analysis, technical studies confirm the utility of CoT%C as a market timing device. Over the 13
year period between January 1986 and November 1998, a CoT%C or
CoT DI methodology outperformed a Buy and Hold Strategy for
bonds. Using a methodology similar to that used by Hayes (1994) in
studying the application of sentiment measures as market-timing devices, CoT%C outperformed both a Buy and Hold Strategy for bonds
as well as CoT%S and CoT%ST. My study applied volatility bands to
smoothed versions of the CoT%C, CoT%S, and CoT%ST time series
derived from the CoT Report and used violation of those bands to
generate buy and sell decisions in the US Treasury market over a 13
year period starting in January 1986 and ending in November 1998.
[see Appendix C for a discussion of the test methodology.] The output of this study is summarized in Table 1.
.50 Std
.75 Std
1.0 Std
.75 Std
1.0 Std
CoT%C
46.1%
53.5%
41.4%
3.2%
3.7%
2.9%
CoT%S
25.9%
26.0%
21.7%
2.0%
2.0%
1.7%
CoT%ST
19.0%
43.6%
39.5%
1.5%
3.1%
2.8%
29.6%
.50 Std
BIOGRAPHY
Sal Greco, CFA, CMT is a Director of Fixed-Income Investments for the Metropolitan Life Insurance Company. He is the
Senior Market Strategist for the Department's Cross Sector Relative Group which addresses issues of asset allocation and relative
value across the company's portfolios. He is responsible for providing both fundamental and technical analyses on the domestic fixed-income markets, in addition to providing both shortand longer-term interest rate and yield curve forecasts for the
trading and portfolio management units of MetLife.
Mr. Greco is also the President of SG Research & Strategy,
L.L.C., a corporation he formed in 1995 to provide consulting
services, market analyses, and trading strategies to individuals
and institutions.
2.2%
Despite the usefulness of a CoT%C and TAbs Net Position methodology over the last several years, the methodology was misleading
at critical points in the 1990-93 Bull Market in bonds, and the Bear
Market which followed in 1994. During the former, the CoT% indicator kept one out of the Bull Market for a continuous 23 month
period. During the latter, it provided a timely sell signal early in 1994,
but reestablished an errant long position for six months in one of
the worst Bear Markets in a quarter century. This can be seen in
Exhibits C-2 and C-3 of Appendix C.
CONCLUSIONS
The analysis of OI under a Money Flow framework derives vital
sentiment information directly from the positions of participants in
a market. This internal, market-generated expression of sentiment
can be used in conjunction with the more popular survey-based, external sentiment measures to gain a more thorough understanding
of the underlying dynamics of the market and its price action.
Used in conjunction with traditional charting techniques, these
methodologies can add significant confirmation to an independent
price action analysis, while providing a quantifiable advantage for
market-timing purposes in the US Treasury market.
MTA JOURNAL
Spring-Summer 2001
15
Appendix A
Figure A-1
Table A1: Decomposition of CoT Report
LARGE
SMALL
Decomposition
The CoT Report decomposes OI into Total Gross Positions (long
and short) held by Reporting and Non-Reporting Accounts. All
clearing members of the exchange, futures commission merchants,
and foreign brokers on the exchange are required to make daily reports to the CFTC showing each of their customer accounts positions on their books, that in any contract month of a commodity,
exceeds the reporting level. If the accounts position does not exceed the reporting level, the aggregate positions of the customer accounts are classified as Non-Reporting for CoT purposes. If they exceed the reporting level, account positions are reported under the
Reporting categor y. [The current reporting level for Bonds is 500
contracts.]
The Reporting Categor y is further divided into Non-Commercial and Commercial categories. Non-Commercials, more commonly referred to as Speculators, are defined as accounts who have
positions in excess of the reporting limit, but do not take positions in
the commodities or securities underlying the futures contract. Commercials, more commonly known as Hedgers, are defined as accounts
whose positions exceed the reporting limit and take positions in the
cash commodities or securities underlying the futures contract. The
Non-Commercial category also contains the positions of Spreaders,
accounts that have offsetting (long and short) positions in different
contract months and do not hold positions in the cash commodity
underlying the futures contract.
The CoT Report provides a decompostion of gross open positions
(long and shorts) on a total basis as well as into each of these catego-
MTA JOURNAL
LONG
SHORT
NET
TOTAL
Comml
424,421
506,241
(81,820)
930,662
Spec
92,484
36,718
55,766
129,202
(NRPT)
91,287
65,233
26,054
156,520
Spreader
12,650
12,650
620,842
620,842
163,640
Open Interest
620,842
608,192
The definitions and terminology of the CoT report can be confusing, especially as released by most data vendors. It is more easily
understood in the context of Table A1, which re-orders the data provided in the raw report of Exhibit A1. Table A1 removes the Spreader
category from within the Non-Commercial category and shows it as
its own category. Representation of the CoT data in this form allows
greater clarity and is the basis for reformulation of the CoT data that
underlies the methodology and indicators I present. Table A2 more
clearly represents the percentage-based measures of positioning in
the CoT Report.
Total OI is defined as one-half of all open positions outstanding.
Thus, Total OI can be derived by summing all open long positions or
all open short positions in the market. As evident from Table A1,
total OI is equally represented by the sum of gross long positions or
gross short positions. By definition, the spreader component has an
equal number of open contract positions on the long and short side
of the market.
Spring-Summer 2001
16
Small
%Long
%Short
Comml
69.8%
83.2%
Spec
15.2%
6.0%
85.0%
89.3%
NRPT
15.0%
10.7%
100.0%
100.0%
Interpretation
The conventional wisdom offered by Belveal (1985) and Jiler (1985)
suggests the LT component of the CoT will outperform the ST component; and that Comml can be expected to outperform Specs. Rather
than address interpretation from the LT v. ST or Comml v. Spec division, it is more valuable to do so from a more generic strong-hand
vs. weak-hand perspective. This avoids the differences between
Comml and Specs that invariably occurs across markets.
Commls represent the strong hands of the market versus the weak
hand of the Spec or NRPT Trader, due to their superior market experience and financial strength. Further, the Comml has qualities
of judgment, experience, and adequate capitalization (Belveal, p
42) which makes him/her a strong hand.
Specifically, Belveal cites many factors that allow Commercials or
Hedgers to represent and maintain the strong hand in the commodity futures market. Because they are not limited by position limits
and have business requirements and the financial wherewithal to buy
or sell in multiples of contract size or market depth, the commercials
are the 100 pound gorilla. The commercial/hedgers proprietary
interest in the underlying commodity allows him/her to conduct defensive buying or selling when price changes appear to threaten his/
hers interests. Finally, the necessity of having to transact in all market conditions and all price ranges makes the Commercial/Hedger
one of the best analysts his/hers market could have. Commercial
players are always faced with the need to sell in low-priced markets
and buy in high-priced markets. In fact, they are often forced to do
one only days or weeks after doing the other. Further, the on-going
nature of hedging operations/activities among commercial accounts
attest to their profitability. If it were otherwise, hedging programs
could not be justified and would rightly be abandoned by the Commercial players inside a market.
Small Traders (NRPT) and Specs are undercapitalized and lack
the information and experience of the larger Commercials. Belveal
sums up their weak-handed nature quite simply: The weak hands
are weak not because they are likely to be wrong in their judgment,
but because they perform so badly in the face of vicissitude. (p. 151)
Such interpretations of positioning among categories of traders
give the decomposition and derivations of the CoT Report a logical
and applicable anecdotal basis for application.
Open Interest =
Comml Long + Spec Long + NRPT Long + Spread Long
or
Comml Short + Spec Short + NRPT Short + Spread Short
Spread Adjusted OI =
Comml Long + Spec Long + NRPT Long
or
Comml Short + Spec Short + NRPT Short
or
Total Open Interest - Spreading OI
Total Absolute Net Positions=
|Net Comml Pos| + |Net Spec Pos| + |Net NRPT Pos|
Fourth, because the sum of total Net Positions must equal zero,
the sum of the Net Position of any two of the trader categories must
be equal in magnitude but opposite in sign to the third category.
This is a corollary to the tenet that Net Open Positions must sum to
zero. However it is often the case but not the rule that the same
two categories will offset the third category especially when Total
Absolute Net Postions are high. For example, 63% of the 482 weekly
observations from the CoT Report for Bonds, Specs were aligned on
the same side of the market (short v. Long) with NRPT.
MTA JOURNAL
Spring-Summer 2001
17
Appendix B
Appendix C
TESTING METHODOLOGY
The identity properties of the CoT Report assure that the sum of
two of the three components (Comml, Spec, ST) will be equal in
magnitude but opposite in sign to the third component.
In order to assess the utility of CoT derived series as market-timing tools, I conducted a study using a methodology similar to that
used by Hayes (1994). [Hayes used weekly Market Vane Sentiment data
from three markets (Gold, Bonds, and Stocks) to generate Buy and Sell signals
in - and across - those markets to determine if an active contrary opinion
investment style outperformed passive Buy and Hold Strategies.] My objective was to determine if an active contrary investment style using the
3 component series I derived from the CoT Report for Bonds
(CoT%C, CoT%S, CoT%ST) outperformed a passive Buy and Hold
strategy for Bonds over the preceding 12 years, and whether one series proved superior to another.
The data requirements for this study were quite simple. The
CoT%C, CoT%S, and CoT%ST series were derived from CFTC data
downloaded from Bloomberg and the CFTC database. The data is
clean and reliable. Daily closing prices for the nearby Bond Future
were taken directly from the Chicago Board of Trade (CBOT) database and from Bloomberg. That information was reviewed and found
to be clean and reliable. Weekly close observations were chosen so
that the pricing day corresponded to the last day covered by each of
the CoT Reports, or to the day closest to the last day covered by the
CoT release. There was no constraint that prices be executable,
since the goal of the study was to determine whether a contrary relationship between Bond prices and CoT Position data existed. The
construction of a trading system was not a goal, so the constraints of
trading system design were not applied.
First, a smoothed version of all three CoT indicators was obtained
by deriving 13-week moving averages of the raw CoT%C, CoT%S,
and CoT%ST time series from the CoT data. Second, volatility bands
were generated at three standard deviation magnitudes (.50, .75, 1.0)
around a 13-week moving average of each of the originally smoothed
data. Third, buy and sell signals were generated only after the
smoothed data moved through a band, reached an extreme, and then
reversed back through the same band. Exhibit C-1 shows the smoothed
data series for CoT%C and the volatility bands generated at a 1.0 Std
level.
Since the CoT%C sentiment series represents strong hands or
smart money, a buy signal is generated when CoT%C moves below the
upper band after violating it on a move to an extreme positive reading.
A sell signal is generated when CoT%C moves above its lower band
after violating it on a move to an extreme negative reading.
The CoT%S and CoT%ST series represent weak hands or dumb
money. Therefore, a buy signal is generated when the smoothed CoT%S
and CoT%ST series move above the lower volatility band after violating
it on the way to a an extreme negative reading. A sell signal is generated
when the smoothed series moves below the upper band after violating it
on a move to an extreme positive reading.
The assumption underlying the model is simple. $1.0 mm is available on 1/2/87. A Buy and Hold strategy is represented by purchase
of a 1 mm Par Bond at the closing price of that day, held over the
entire period, and valued at the closing price on the last day of the
study. The active management style was driven by Buy and Sell signals generated from our three CoT derived sentiment series and their
volatility bands. The $1.0 mm opening balance remained in cash
until a Buy signal was generated by the model. The model was in
Neutral mode at the start of the study period.
Exhibit C-2 displays the price action of the Nearby Bond Future
along with the Buy and Sell signals generated by the CoT%C series at
MTA JOURNAL
Spring-Summer 2001
18
Exhibit C-1
the 1.0 std band level. It outperformed a Buy and Hold Strategy over
the the 12 year period by 1180 bps, or approximately 290 bps on an
annually compounded basis. Table C-1 summarizes the output for all
three data series at all three std levels. [This table also appears in the
body of the text as Table 1.]
An active CoT%C strategy would have outperformed a Buy and
Hold Strategy over the 1986 to 1998 period, suggesting that CoT%C
is a valid sentiment indicator, and that a Money Flow framework for
OI Analysis can improve market timing decisions in the Treasury market. Further, the CoT%C appears superior to both the CoT%S and
CoT%ST series. This supports the contention that CoT%C - and its
many equivalents - is the broadest and most accurate decomposition
of the CoT data.
However, two things should be noted in regard to the application
of this methodology. First, the output exhibits significant variance
Std magnitudes. In fact at higher levels of Std (>1.0), none of the
series returned profitable results versus a Buy and Hold strategy. This
is a result of timing sensitivity. Clearly, marginal differences in timing of Buys and Sells can change the results dramatically. Since Bond
prices can change dramatically week to week, small differences between signals and market highs and lows can inject a considerable
amount of variance into this type of study. Exhibit C-3 lists the Buy
and Sell signals for each series at each Std level.
Exhibit C-2
.50 Std
.75 Std
1.0 Std
.50 Std
.75 Std
CoT%C
46.1%
53.5%
41.4%
3.2%
3.7%
2.9%
CoT%S
25.9%
26.0%
21.7%
2.0%
2.0%
1.7%
CoT%ST
19.0%
43.6%
39.5%
1.5%
3.1%
2.8%
29.6%
1.0 Std
2.2%
Exhibit C-3
MTA JOURNAL
Spring-Summer 2001
19
INTRODUCTION
In Trading For Profit With Precision Timing, Don Vodopich enhances the
analysis of Elliott Waves by attaching Gann lines to wave starting points.
With this approach, Vodopich studies a variety of futures markets in the shortterm as is appropriate for trading. The purpose of this research is to apply
Vodopichs technique to the long-term study of the stock market, specifically the
Dow Jones Industrial Average (DJIA), which was the focus of Elliotts work.
The period covered is Supercycle Wave (V) as counted by Frost & Prechter,
which began in 1932. The Vodopich analysis breaks down the present cycle of
Cycle-degree (i.e., Supercycle Wave (V)) as well as certain Primary, Intermediate, and Minor cycles within it.
BACKGROUND
Vodopich argues that the slope of the line from the start of an
impulse wave to its end can be known in advance because the slope
is particular to the tradable.1 For example, the slope of an impulse
wave of several months length in T Bond futures always has a slope
of .10 points (about 3 ticks) per trading day. Such a line is a Gann
line since unlike a trendline its slope can be known at its starting
point, but like a trendline it identifies support and resistance. Chart
1 demonstrates how such a Gann line attached to an impulse wave
might work in theory. The Gann line would indicate support up
through the top of impulse wave 3, would be firmly broken by wave 4,
and would indicate resistance thereafter up to the end of the impulse wave, the end of wave 5.
Chart 1
Gann Lines Drawn Onto Elliott Impluse Wave
Chart 2
T Bond Futures Contract from April '97 to January '98
MTA JOURNAL
Spring-Summer 2001
20
start of wave 3 at 107.75 on 5/27. Inspecting the chart, one can see
that the .10 line did reveal support up through the wave 3 top on
8/1, capturing the late June correction. Note that a line drawn from
the beginning of wave 1 with the same slope would have sliced through
the ensuing rally, rendering the Gann line useless.
Wave 4 of this rally ended at the 8/26 low of 111.438, firmly breaking the .10 line. Both during wave 4 and after, this same line identified resistance until January 98. The second Gann line begins the
3rd wave within wave 5, at the 9/11 low of 111.938. It has a slope of
.15 points per trading day, 1.5x the slope of the line attached to the
parent wave. The .15 line indicated support through late November 97. After that line was firmly broken by the 4th wave within wave
5, the last Gann line was drawn from the low of 117.78 1 on 12/9/97
(using the 3/98 contract). The slope chosen for this line was .20
points per day, 2x the slope of the original Gann line and 1.33x the
slope of the second line. This slope is chosen in part because it directs the line to the intersection of the first and second Gann lines.
This intersection, a Vodopich convergence, at 124.5 in mid-January
1998, is near the intraday high of 124.28 on 1/12/98 (using the 3/98
T Bond futures contract), which remained the high for several
months.
This demonstration of Vodopichs integration of Elliott and Gann
supports its theoretical argument, that an Elliott impulse wave is associated with a Gann line whose slope has worked for the tradable
in the past1; that Gann lines of component waves have slopes that
are round multiples of the slope of the parent wave; that these Gann
lines indicate support through wave 3 and resistance thereafter; and
that the Vodopich convergence, the intersection of the Gann lines,
identifies a point in time/and or price (both time and price in this
case) at which the reward/risk ratio drops substantially. Appendix 2
contains the dates and prices of the Elliott waves discussed above with
relevant Fibonacci ratios.
Readers familiar with the T Bond action in 1998 know that after
the January high, T-Bonds corrected until April 98 and then rallied
to a historic high in October. This price movement raises the question whether there was a larger cycle operating which this Vodopich
analysis did not capture. The issue of a hierarchy of cycles is central
to the long-term Elliott Wave analysis of the stock market, the focus
of this research and of Elliotts original study.
MTA JOURNAL
Spring-Summer 2001
21
Chart 3
The Present Minute Cycle (Minor Cycle Wave 5)
Dow Jones Industrial Average from 1994-1997
Price firmly breaking the Gann line (as the July 96 correction
broke the 34% line) is reason to suspect a wave 4 correction. Upon
resumption of the uptrend, one draws the second Gann line, repeating the evaluative process that led to the first. Now, however, during
the wave 3rd wave within the 5th wave, one watches for the second
Gann line to indicate support while the first line indicates resistance;
and this happened in Minor Cycle Wave 5 from September 96
through March 97. With the third Gann line, drawn after the second is broken, one is approaching either the end of the impulse wave,
or, as is the case in Chart 3, a significant change in trend, reducing
the reward/risk ratio. The third Gann line confirms the intersection
of the first two; and as price approaches the Vodopich convergence,
one can adjust ones equity exposure for the expected higher volatility and lower reward/risk ratio.
The Vodopich convergence is not a reversal signal. As has been
discussed in the T Bond futures represented in Chart 2 as well as the
U.S. equity bull markets shown in Charts 3 and 5, more often than
not, the trend resumes after a Vodopich convergence, albeit with
greater volatility and after some consolidation. This likelihood points
out a limitation in the authors use of Vodopichs integration of Elliott and Gann, since the addition of Gann lines leaves part of the
impulse wave unmeasured. An addendum to this paper examining
the price action from August 97 to the present addresses the question of how to analyze such price action.
In summary, Chart 3 demonstrates a Vodopich analysis of a longterm (32 month) Elliott impulse wave of the Dow. The first and second Gann lines, at 34%/yr. and 51%/yr., indicate support through
the 3rd wave of their cycle and resistance thereafter. However, the
third Gann line, the 102% line, performed differently from the
first two, indicating support through the 3rd wave but never resistance. All three lines intersected in late July 97, near the 8/7/97
high of 8299.49. And while this Vodopich convergence did not mark
the absolute high, it did indicate the time and price after which price
movement became more volatile. The Dow did not break the 8299
high for six months, in February 98; and from August 97 through
August 98, the Dow suffered two significant corrections, -16% in
October 97 and -21% in July/August 98.
The discussion of Minor Cycle Wave 5 is also a practical example
of how one applies Elliott and Gann to long-term stock market impulse waves. Early into what one believes is a 3rd impulse wave, early
1995 for the cycle in Chart 3, one draws a Gann line whose slope is a
round multiple of the slope of the parent-wave Gann line; at the time,
this is an academic exercise since the Gann line does nothing to establish that the cycle has begun its 3rd wave. The inability of a Gann
line to contribute to the assessment of a 3rd wave is no drawback
since 3rd waves are the easiest of all waves to spot. In the words of
Frost & Prechter, Third waves are wonders to behold. They are strong
and broad, and the trend at this point is unmistakable. Increasingly
favorable fundamentals enter the picture as confidence returns . . . .
Such points invariably produce breakouts, continuation gaps, volume expansions, exceptional breadth, major Dow Theory trend confirmations and runaway price movement.4
Relying on other measures, then, to validate the wave 3, one can
experiment with Gann lines of two or three different slopes (e.g.,
1.5x, 2x, 2.5x the slope of the parent wave) to find the line that best
indicates support. In the latter half of a 3rd wave, in late 1995 for
Minor Cycle Wave 5, one should expect to find the single Gann line
that best indicates support through the 3rd wave. The analyst must
be careful not to create a Gann line that is too tight, whose slope is
too steep; such a line probably belongs to the 3rd wave within wave 3,
not wave 3 overall. To check that the Gann line chosen is proper
and not too steep, the author draws a price channel, using the Gann
line as its lower boundary. The usefulness of this technique is based
on the authors observation that price channels are common and
easy to see in 3rd waves.
MTA JOURNAL
THE CURRENT PRIMARY C YCLE (CYCLE WAVE V) THE BULL MARKET FROM 1974
Chart 4 plots monthly bars of the DJIA from the 1974 low through
March 1998 and indicates the particular wave count for this Primary
cycle which the Vodopich analysis suggests. For reference, note that
Chart 3 (covering the period from 1994 onward) fits in the upper
right hand corner of Chart 4. Note also that the 34% line in Chart
4 is the same line of the same slope from Chart 3. The Vodopich
analysis of this period leads to a different wave count than that suggested by Frost & Prechter5; and that wave count, in turn, leads to
two important conclusions. The Vodopich analysis suggests first that
the entire period from 1982 to the present is Primary Cycle Wave
((3)) and second that the 1990 high, not the 1987 high, was the top
of Intermediate Wave (3) within ((3)). The argument here for the
revised wave count is one of fit rather than proof. The explanation fits the market action to date. However, it does not provide a
perfect Elliott wave count, nor is it the necessarily the count that
would have presented itself at various times in the past.
In this count, the Primary Cycle (Cycle-degree Cycle Wave V) begins at the 12/9/74 low; and the first Gann line, with a slope of 12.75%,
begins at the 8/31/82 low of 769.98, which in this count represents
the start of Primary Wave ((3)). Here again, the integration of Elliott and Gann is iterative; neither the wave count nor the Gann line
were fixed before adding the other. The choice for the Gann-line
slope and the start-date for drawing it (with the wave count implied
by this choice) are made in an effort to find the best wave count and
Gann line for the price action that followed.
What is fixed in a Vodopich analysis is that the Gann lines slope
must be a round multiple of the Gann-line slope of the parent wave
(12.75%/yr = 1.5 x 8.5%/yr.) and that the Gann line indicates support until it is firmly broken by the 4th wave in the cycle, wave ((4))
in this case. As one can see from Chart 4, the 12.75% line has
remained consistently beneath the market lows after 1982, i.e., the
lows in 1984, 1987, 1990, and 1994.
Spring-Summer 2001
22
Chart 4
The Current Primary Cycle
(Primary Waves ((1)) through ((3)) Cycle-degree Cycle Wave V)
Dow Jones Industrial Average from 1974-1997
Table 2
Projections of Gann Lines to August 7, 1997 and
Comparison with Actual High for Dow Jones Industrial Average
MTA JOURNAL
Chart 5 plots monthly bars of the DJIA from 1942 to 1972, the
bull market between World War II and the Vietnam War; and the
Vodopich analysis confirms Frost & Prechters wave count for this
period (Cycle Wave III).7 Wave III of Supercycle Wave (V) began at
the 4/28/42 low of Dow 92.92; and from here the Gann line with the
slope of 8.5%/yr. is drawn. This line clips the 6/13/49 low of 161.5,
which marks the start of Primary Wave ((3)) within Cycle Wave III.
And from this point, a Gann line is drawn with a slope of 12.75%/yr.,
1.5 x 8.5%/yr., the slope of the parent wave. The 12.75% line indicates support during Wave ((3)), capturing the pullbacks in 1953
and 1957, until it is broken in the early 1960s. In this way, the 12.75%
Spring-Summer 2001
23
line validates Frost & Prechters count of the Wave ((3)) top in 1958.7
The second Gann line for Cycle Wave III is drawn from the start
of Wave (3) within ((5)), at 558.06 on 10/32/62. Its slope of 25.5%/
yr. is 2x the slope of the Gann line of the parent wave, and 3x the
slope of 8.5%, the slope of the grandparent wave, Cycle Wave III.
This second Gann line behaves as a third Gann line often does, not
indicating resistance once broken; and no third Gann line emerges
from the analysis of this period to confirm the intersection of the
first two. The first two lines do intersect in early 1966 around Dow
1250, and this marked Frost & Prechters Cycle Wave III top in time
but not in price.
Here again, the Vodopich convergence does not signal a reversal
but rather the beginning of a period in which volatility increases substantially with some advance in price. The 1973 Dow peak of 1051.7
exceeds the 1966 peak by 5.7%; but in between the peaks the Dow
suffers two corrections, -25% in 1966 and -36% in 1968-70. Frost &
Prechters use of complex wave patterns to analyze this period is taken
as support for the argument that the portion of a trend after the
Vodopich convergence is inherently more difficult to analyze with
the Elliott Wave. To analyze the time from 1966 to 1982, Frost &
Prechters original (1978) count labels the period from 1966 to 1974
as an expanding triangle; but in 1982, Prechter favored a doublethree wave count, ending Cycle Wave IV in 1982.5
An addendum to this paper analyzes the Dow following the
Vodopich convergence in August 97. In this real-time post-convergence analysis, the author will suggest how to deal with such a period.
Chart 6 plots quarterly (3 month) bars of the Dow Jones Industrial Average from 1910 to the present along with three Gann lines
and Cycle Wave turning points. The first two Gann lines, with slopes
of 8.5% and 12.75%, are identical to the lines with the same slopes in
Charts 5 and 4 respectively. As one can see in Chart 6, the Gann lines
applied to the Dow from World War II onward indicate support and
resistance in the same manner as Gann lines in Charts 2 through 4.
The first Gann-line, the 8.5% line, indicates support through Cycle
Wave III; this line is firmly broken by Cycle Wave IV; but the line does
not indicate any meaningful resistance until the summer of 1997, as
has been discussed previously.
MTA JOURNAL
Spring-Summer 2001
24
SUMMARY
The purpose of this research has been to apply Vodopichs integration of Elliott and Gann to the study of long-term impulse waves
in the Dow Jones Industrial Average. The effectiveness of such an
analysis can be seen by comparing Charts 1 and 6. Chart 1 shows how
Gann lines should help a wave count; and Chart 6, the DJIA after
World War II with Gann lines based on percentage slopes, delineates
a credible wave count for Supercycle Wave (V) and targets its top.
Drawing Gann lines from Elliott Wave starting points can help select
the best among alternative wave counts; but it does not alter the iterative, trial-and-error process of establishing the count in the Elliott
Wave. As Frost & Prechter state, one can use the market action to
confirm the wave count as well use the wave count to confirm the
market action.11 Plotting a Gann line from the suspected start of a
3rd wave is an academic exercise. The Gann line itself does nothing
to establish that the cycle has begun a 3rd wave; and only if price
finds support at a Gann line during a 3rd wave can one conclude that
the line is appropriate.
The Gann lines contribute to the evaluation of the cycle only during the latter stage of the 3rd wave and during the 4th and 5th waves.
Evidence of the end of the 3rd wave and the start of the 4th wave
correction comes from price breaking below the first Gann line of
the cycle. During the 5th wave, one repeats the trial-and-error process to create the second Gann line, using the slope of the first Gann
line as an aid in choosing the slope of the second. The movement of
price in between the first (now resistance) and second (support) Gann
lines during a 5th wave gives the analyst confidence in the 5th wave,
which the author has found to be more volatile than the well-channelled 3rd wave, thereby obscuring the wave count. Finally, at the
price and/or time target indicated by the intersection of the two Gann
lines (perhaps with the confirmation of a third line), the analyst can
reduce market exposure, both confident that the predictable bulk of
the profit has been achieved and wary of the much greater volatility
that tends to follow a Vodopich convergence.
A general caution must be added about the reliability of the Gann
lines in the way they indicate support and resistance. Price may overthrow (rise above) the Gann lines especially near the top of the cycle
due to the momentum built up through the 5th wave; both Frost &
Prechter12 and Vodopich13 have observed such throwover. Conversely, price may fall away after breaking below the third Gann line
in an impulse wave so that the third line plays no resistance role at
all. And especially in smaller impulse waves, one may not be able to
see a third Gann line. However, to the extent that Gann lines contribute to finding the correct wave count, they sensitize the analyst to
other market measures signalling a reversal.
Finally, the Supercycle Wave (V) target of 29,000 in 2012 must be
considered preliminary since it lacks the confirmation of the third
Gann line. As we near the top, however, analyzing component waves
with Vodopich's approach will help bring that Supercycle high in the
Dow into focus.
ADDENDUM - (4/30/99)
This addendum covers the DJIA after the Vodopich convergence
in August 97 (see Chart 3). On a smaller degree, smaller by one
Elliott-Wave degree in fact, this period is like the years from 1966 to
1973 (see Chart 5). What emerges from analyzing these two periods
is that the price action up to a Vodopich convergence fits well into the
Elliott Wave hierarchy, but that after the intersection of the Gann lines,
price does not fit neatly into a single cycle that itself fits inside a larger
cycle. The deep retracements of a post-convergence period undercut
the search for a single impulse wave with inter-related Gann lines.
MTA JOURNAL
Spring-Summer 2001
25
Appendix 1
Gann Line Slopes of Various Tradables
Appendix 2
Prices, Dates & Fibonacci Ratios for T Bond Future Contract from
April '97 to January '98
MTA JOURNAL
Spring-Summer 2001
26
Appendix 3
Prices, Dates & Fibonacci Ratios for Minor Cycle Wave 5
BIBLIOGRAPHY
BIOGRAPHY
This research, submitted to the MTA in May 1999, fulfilled
the authors Phase III requirement of the CMT Program. Mr.
Hulton is indebted to his CMT Mentor, Gurney Watson, for his
guidance and support.
Mr. Hulton is presently Director of Trader Services at the
Electronic Trading Group, L.L.C.
ENDNOTES
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
Vodopich, p. 59.
Frost & Prechter, p. 25.
Ibid, p. 71.
Ibid, p. 80.
Ibid, p. 283.
Barrons, 6/29/98, p. 46 (Inter view with Mark Perkins).
Ibid, p. 282.
Dent, p. 294.
Frost & Prechter, p. 67.
Pring, p. 255.
Frost & Prechter, p. 84.
Ibid, p. 73.
Vodopich, p. 55.
Vodopich, p. 47.
Frost & Prechter, p. 31.
MTA JOURNAL
Spring-Summer 2001
27
With earth's first clay they did the last man knead,
And there of the last harvest sowed the seed.
And the first morning of creation wrote
What the last dawn of reckoning shall read.
Omar Khayym
INTRODUCTION
It strikes me that a philosophical discussion of the Elliott Wave Principle is
a worthwhile and relevant pursuit, especially since some of its most noted
practitioners, such as A. J. Frost and Robert Prechter, do not hesitate to discuss philosophical issues in their writing nor do they fail to quote various
philosophers in connection with their own advocations. I believe that Frost
was the first to specifically articulate the connection between philosophy and
the Wave Principle as evidenced by his observation that [i]t is possible to
read into stock market behavior a philosophical significance under the basic
tenets of the Elliott Wave Principle. 1
There is no shortage of philosophical issues that can be discussed in relation to the Wave Principle, but the limitations inherent in this paper force me
to zero in on a select few. As such, I have decided to examine determinism and
specifically whether it should be said from a philosophical point of view that
the Wave Principle is inherently deterministic. This consideration also seemed
to have occurred to Frost: It is an open question whether or not man is a
puppet on a string. In the short term he is not, but over the longer period he
may be2 and Prechter, the Wave Principle form shows that a collective system
is...deterministic.3
This paper is not about the persuasiveness of any particular form of determinism nor the validity of the Wave Principle. Instead, my intent is to investigate the compatibility, or lack thereof, between the two. Deciding whether the
Wave Principle is deterministic is a challenging task since it requires a metaphysical examination of a theory that by its nature, was not directly intended
to address metaphysical issues. There is an important difference between the
determinism of a particular theory and the more enveloping, less precise notion that the world is itself deterministic. This latter view embraces a much
bolder metaphysical view and while it can be supported, it requires more than
just the consideration of the determinism of a particular theory.
As is the case with any philosophical undertaking, formulating a precise
definition of the beginning precepts is a daunting, yet not insurmountable,
task. However, a discussion of the semantics of the determinism is not the
primary focus of this paper, and so, I will use determinism in a general philosophical way while admitting at the outset that there are various versions of
determinism that I will omit due to the limitations of this paper. For the purposes of this paper, I will examine scientific determinism, and show why I
believe that a case can be made that it is compatible with the Wave Principle.
Chart 1
MTA JOURNAL
Spring-Summer 2001
28
tory, the great course of human affairs, has not been the result
of voluntary efforts on the part of individuals or groups of individuals, much less chance; but has been subject to law.8
Cheney's view strikes me as consistent with the Wave Principle since
they both advocate that there is a definite pattern of development in
the apparently jumbled story of human history.9 Further, they both
argue that a system's general characteristics can be said to be determined, at least in part, by the structures and characteristics of the
constituents of those systems.
Every human event has a definite place in an unalterable and consistent structure of progress and regress as each society passes through
a defined series of antecedent changes in order to achieve a subsequent stage. Though individuals are agents that seem to bring about
the specific events of social history, they are simultaneously the instruments by which certain laws (Elliott waves) and mathematical
principles (Fibonacci) relating to the character of social action become manifest.
Both Cheney's view of history and the Wave Principle share the
common premise concerning the impotency of deliberate individual
actions to alter the course of social trends. Or as Voltaire poignantly
said in the passage on "Destiny" in his Philosophical Dictionary, "Everything is performed according to immutable laws...in spite of you."10
This line of thought argues that historical changes are the products of deep-lying forces which conform to fixed, although not always apparent, patterns of development in mass psychology. Didier
Sornette and Anders Johansen of the Niels Bohr Institute in Copenhagen provided evidence for the existence of a macro intelligence
when they wrote in 1997:
DETERMINISM
This metaphysical principle has been understood and assessed in
various ways over the centuries. Discussions concerning determinism are often challenging due to the concept involved: what "determinism" means. Since the 17th century, it has commonly been accepted as the doctrine/theory that all human thought, action or event
is caused entirely by preceding events. This is to say that all physical
events and human actions are determined by antecedent factors.
Philosophers distinguish between hard determinism (e.g. necessitarianism,4 fatalism5) and soft determinism (e.g. compatibilism6 , libertarianism7 ).
In addressing the validity of determinism, philosophers have often looked to science and scientific theory to represent the best guide
to the truth of determinism. Many philosophers have discussed determinism in light of opinions about metaphysical topics such as free
will or God. Historically, the principle of determinism applied to
both secular and theistic philosophers. While some thinkers, like
Immanuel Kant, discussed determinism in terms of the science of
the day, others made it part of their philosophy of nature.
Various writers have often referred to determinism, causality and
natural law as if they were synonymous. They are connected, despite
the fact that they are not fully equivalent and the differences between
them can be clarified if one wanted to make his/her analysis meticulous enough. For the purposes of this discussion, however, I do not
find it necessary to attempt to sharpen distinctions to the point of
emphasizing all the recognizable differences between the terms. Instead, I shall use them in an essentially common-sense fashion.
MICRO OR MACRO
In the early 1930s, accomplished historian Edward Cheney studied various historical events in relation to the seemingly influential
actions of certain famous figures. After examining the ostensibly decisive effect exercised by individuals at the time and the role they
played in helping to bring about these historical events, Cheney concluded:
These great changes seem to have come about with a certain
inevitableness; there seems to have been an independent trend
of events, some inexorable necessity controlling the progress
of human affairs. Examined closely, weighed and measured
carefully, set in true perspective, the personal, the casual, the
individual influences in history sink in significance and great
cyclical forces loom up. Events come of themselves, so to speak;
that is, they come so consistently and unavoidably as to rule
out causes not only of physical phenomena but voluntary human action. So arises the conception of law in history. His-
MTA JOURNAL
Spring-Summer 2001
29
SCIENTIFIC DETERMINISM
MTA JOURNAL
1.
2.
3.
4.
Spring-Summer 2001
30
Chart 2
MTA JOURNAL
inventions and creation and that the emergence of these are not predictable. No one, for example, could have predicted Count Basie's
April in Paris or Darwin's Theory of Evolution. What the Wave Principle
can and does do, however, is ascribe the deterministic patterns that
allow for the favourable social conditions that are conducive to both
the undertaking of research and the acceptance of innovative discoveries and creations.
Whether specific events will take place is a question of probability
since the Wave Principle only dictates society's coming character
changes, but not necessarily specific events. This is to say that the
aggregate social mood fashions the character of history, not the specific manifestations.
Individual occurrences and events do not follow deductively from
the wave pattern. While the environment is certain and determined
by the wave pattern, the individual occurrences within that environment are based on probability.
Notice on Chart 3 that, for example, according to this particular
Wave count, after the Supercycle Wave 2 low in 1859, it was predict-
Spring-Summer 2001
31
Chart 3
CONCLUSION
During this paper, I have not attempted to defend either the validity of the Wave Principle nor
scientific determinism. My intention throughout
has been to determine whether the Wave Principle was compatible with this version of determinism.
I have drawn a number of conclusions from
this work. The Wave Principle seems to be consistent with scientific determinism. Aggregate
social mentation and the fundamental fractal pattern of the Wave Principle moves according to a
determined yet robust path that is based on natural laws and antecedent conditions. This is not to
say that one can identify determinism or the Wave
Principle with the potential to predict with unlimited accuracy. At the same time, however, they
are directed towards discovering regularities in
the operations of systems that will empower one
with the ability to formulate various rules that allow for the prediction or retrodiction of the occurrence of events.
I believe that based on the determined path of the Wave Principle, it is reasonable to say that we can predict the future behavior of
a large body of people but not the individuals that compose that body.
In other words, explanations of aggregate social mood based on wave
patterns have the structure of a straightforward deductive argument
while individual actions do not.
EPILOGUE
Due to the inherent limitations of this paper, I have had to limit
the scope of my examination to specific areas. There remain, however, many other philosophical areas left to investigate.
As we move into the 21st century, technical analysis continues to
grow in reputation and stature, so much so that it is slowly being
accepted by mainstream academia. Hence, a philosophical look at
some of the techniques practiced and the beliefs held by technicians
is an area that I believe holds great potential. It is, in Shakespeare's
words, "undiscovered country." If we do indeed learn atop the shoulders of those who came before us, then I believe that some of
philosophy's greatest minds have valuable insights for us as technicians.
When I refer to the dearth in philosophical analysis of the financial markets, I am not referring to Wave Principle specifically (though
there is much more to say about philosophy and Elliott Wave) but
technical analysis in general. For example, the concept of the herd,
central to the technicians' view of market movement and market sentiment, was explored in detail by Nietzsche, Kierkegaard and
Heidegger. Aristotle had a lot to say about man as a "political animal"
and social creature. Thomas Hobbes' famous belief that a state of
nature was a state of war and that the life of man was "solitary...nasty,
brutish and short" certainly seems to ring true with not only the competitive, everyone-for-himself nature of trading, but also the typical
longevity of most traders. There is much we can and should learn
from these thinkers...one step backward, two steps forward.
In short, I believe there is a book waiting to be written. This is a
book that will look at the greatest philosophical minds and explore
Because fundamental Elliott wave patterns are limited in number... and because the continual expansion of degree imparts
uniqueness to every wave, interactive human mentation and
behavior, which produces history, are continually repeated, but
not precisely.23
I have one last point to make. Just as social-mood conditions are
determined but social events are not, the form of the impulsive and
corrective patterns are determined, but the dimensions of the waves
within the pattern are not. This leads to an important point regarding the nature the predictive power of Elliott waves. The general patterns of the Wave Principle are universal in form. But in reality, the
patterns are rarely, if ever, found asserted in precisely the same way
(for example, the length of fifth waves are not exactly the same in
every pattern even though they tend to gravitate towards predetermined Fibonacci projections), analogous to Einstein's famous expression that "so far as the laws of mathematics refer to reality, they are
not certain. And so far as they are certain, they do not refer to reality."
The consequence is that in applying the generalized patterns to a
specific occurrence, there will be some uncertainty concerning
whether the given situation adheres to the generalizations. The result of this is that aside from being unable to give the initial conditions that will result in an exact consequence (the termination of
wave five will end at .618 the net of wave 1-3 if and only if...) one can
MTA JOURNAL
Spring-Summer 2001
32
19
what they had to say about the state of nature, human thought and
our social condition and compare it to what we know and see in the
financial markets. As the layers of market analysis are peeled back
and examined, I have every confidence that one will find the footprints of history's greatest philosophers who, if listened to carefully,
have something important to tell us.
20
21
ENDNOTES
1
2
3
10
11
12
13
14
15
16
19
18
22
Robert R. Prechter, Jr., The Elliott Wave Writings of A.J. Frost and
Richard Russell, (Georgia: New Classics Library, 1998), p. 401.
Ibid., p. 398.
Robert R. Prechter, Jr., The Wave Principle of Human Social Behavior
and the New Science of Socionomics, (Georgia: New Classics Library,
1999), p. 414. Referred to as Socionomics for the remainder of
this paper.
Necessitarianism holds that humans do not have free will and that
actions are entirely determined by antecedent, external causes.
Philosophically, fatalism holds that the suffering and despair are
the inevitable fate of man. It is often used in conjunction with
determinism since it argues that every event is bound to happen
as it does not matter what we do about it. Like necessitarianism,
fatalism denies that human actions have any causal efficacy. A
determinist may believe that a hangover is the effect of a natural
cause, but the fatalist holds that a hangover will occur regardless
of whether one drinks or not.
Compatibilism acknowledges that all events, including human actions, have causes. But it allows for free actions when the actions
are caused by one's choices rather than external causes.
Metaphysically, the term libertarianism refers to the idea that human beings have free will and thus sees no inherent contradiction
between determinism and the proposition that human beings are
sometimes free agents.
Ernest Nagel, "Determinism In History," Philosophy and Phenomenological Research, Volume XX, September, 1959. p.291.
Socionomics is a field of study coined in Prechter's previously mentioned book, The Wave Principle of Human Social Behavior and the
New Science of Socionomics. It is the examination of social trends
based on the biologically based patterns of fluctuation in collective mood that are formological in that they have consistent Fibonacci-based mathematical properties and produce the Wave
Principle. Prechter argues that the patterning of social mood
guides and influences the character of individual and social behavior resulting in human actions that in turn cause the trends
and events of history.
M. De Voltaire, A Philosophical Dictionary, (London: C.H. Reynell,
1824), p. 385.
Prechter, Socionomics, p. 159.
Robert R. Prechter, Jr., R. N. Elliott's Masterworks, (Georgia: New
Classics Library, 1994), p. 216.
Prechter, Frost and Russell, p. 378.
Ibid., p. 398.
Ibid., p. 403.
Pierre Simon Marquis de Laplace. A Philosophical Essay on Probabilities, (New York: Dover Publications, Inc., 1951), p. 3.
Prechter, Masterworks, p. 216.
This is an epistemological claim, not an ontological one. I do not
believe there is any compelling reason to believe that if the Wave
Principle is true, it would communicate, as Galileo said, "the language in which the Book of Nature is written." Perhaps a language but not the language.
MTA JOURNAL
23
BIBLIOGRAPHY
Bernstein, Peter L. Against the Gods, New York: John Wiley & Sons,
Inc., 1996.
De Voltaire, M. A Philosophical Dictionary, C.H. Reynell: London,
1824.
Dray, W. H. "Determinism in History." The Encyclopedia of Philosophy, Paul Edwards, editor. Volume 2, New York: MacMillan Publishing Co., and Inc. & the Free Press, 1967. Pp. 373-378
Frost, A.J. and Prechter Jr., Elliott Wave Principle. Georgia: New Classics Library, 1998.
Gies, Joseph and Frances, Leonard of Pisa and The New Mathematics
of the Middle Ages, Georgia: New Classics Library, 1969.
Hook, Sidney. Determinism and Freedom in the Age of Modern Science,
New York: New York University Press, 1958.
Marquis de Laplace, Pierre Simon. A Philosophical Essay on Probabilities. New York: Dover Publications, Inc. 1951.
McKeon, Richard. The Basic Works of Aristotle, New York: Random
House Inc., 1941.
Nagel, Ernest. "Determinism in History." Philosophy and Phenomenological Research, Marvin Farber, editor. Volume XX, New York:
University of Buffalo, 1959. Pp. 291-317.
Taylor, Richard. "Determinism." The Encyclopedia of Philosophy, Paul
Edwards, editor. Volume 2, New York: MacMillan Publishing Co.,
and Inc. & the Free Press, 1967. Pp 359-373.
Prechter Jr., Robert R. R. N. Elliott's Masterworks. Georgia: New
Classics Library, 1994.
Prechter Jr., Robert R. The Wave Principle of Human Social Behavior
and the New Science of Socionomics. Georgia: New Classics Library,
1999.
Prechter Jr., Robert R. The Elliott Wave Writings of A.J. Frost and
Richard Russell. Georgia: New Classics Library, 1996.
BIOGRAPHY
Jordan E. Kotick has an Honours Bachelor of Arts degree
with a double major in Economics and Philosophy, a Master of
Arts degree in Philosophy and is a Chartered Market Technician. Jordan was previously employed with CIBC World Markets
as a technical analyst and government bond trader. He is currently Senior Technical Analyst for Elliott Wave International
and Vice President of the Canadian Society of Technical Analysts. He can be reached at: jordank@elliottwave.net
Spring-Summer 2001
33
Under this rule, the risk of being stopped out is highest when a
position is newly put on. Once it is in the money, the risk of being
stopped out is reduced.
The weakness of this stop-loss rule is that it becomes less effective
if positions are marked to market periodically, such that a big loss
could still show up on the monthly (or quarterly) valuations notwithstanding that the cumulative P/L since inception remains
positive.
2) Trailing stop-loss
Similar to the fixed % stop-loss, the trailing stop-loss rules require
investors to pre-specify a maximum loss limit, at which the position would be closed out.
The key difference from the fixed % stop-loss is that the stop-loss
trigger does not apply to the cumulative P&L. Instead, it applies
to the P&L relative to the highest profit achieved. Suppose a 2%
threshold is chosen: then the position would be stopped out once
the mark-to-market P&L drops more than 2% from its highest
point.
Therefore, this stop-loss rule not only seeks to protect capital but
also the profit accrued.
3) Other stop-loss rules
There are many other different kinds of stop-loss rules. These include:
I. Standard Deviation Stop-loss
Instead of defining the stop-loss in terms of a fixed % or dollar
amount, a standard deviation threshold can be used to achieve a
similar objective.
II. Price Level Stop-loss
The stop-loss trigger is set relative to the market price of the instrument traded rather than the P&L. The trigger is usually set at
a price level of technical significance, i.e. a support or resistance
level. This is very popular among short-term traders whose trading activities are driven primarily by technical analysis.
III.Event Driven Stop-loss
For instance, if a trade position is assumed based on the belief
that the central bank will intervene to support a certain level. The
absence of intervention after the breach of the level would trigger a stop-loss.
Each rule has its theoretical appeals, but for the purpose of this
study, we have only tested the fixed % stop-loss and the trailing stoploss rules, which are the most popular and widely used ones. As they
possess most of the basic and important characteristics of stop-loss
rules, the results should have a meaningful degree of general applicability.
INTRODUCTION
The objective of this paper is to study the efficacy of stop-loss mechanisms.
We have back-tested two of the most commonly applied stop-loss rules (fixed %
and trailing % stop-loss) on both a trend-following and a mean-reversion
trading strategy. The results from our tests indicate a significant negative
correlation between the value added by applying these stop-loss rules and the
profitability of the trading strategies tested. In other words, there was a tendency for these stop-loss mechanisms to undermine the performance of the profitable trading strategies and improve that of the unprofitable ones tested.
The results led us to question the expectation that stop-loss will enhance
trading performance over time by reducing the size of losses and argue that
investors should carefully examine (e.g. via back-testing) the impact of stoploss mechanisms on their trading strategies/style prior to adopting them.
In addition, we have discussed several important qualitative attributes of
stop-loss, which we believe should also be considered when assessing the desirability of using stop-loss. These include a reduction in the "risk of ruin,"
dilemma in re-entry and the impact of the "fear of regret" syndrome.
We conclude with a warning against a blind acceptance of the utility of
stop-loss, or even worse, an over-reliance on stop-loss. As indicated by our test
results, stop-loss influences primarily the "pattern" of loss rather than the
"probability" of loss. To most investors, it is the latter which matters. To effectively manage the "probability" of loss, investors should seek to control "risk,"
which is an ex-ante parameter, rather than "loss," which is an ex-post parameter. Risk-control requires a careful management of the size and nature of
exposures. Stop-loss, on its own, is not sufficient to achieve effective risk-control. An over-reliance on stop-loss may even generate a false sense of security,
with undesirable consequences.
MTA JOURNAL
Spring-Summer 2001
34
Equity:
FX:
S&P 500
(USA)
FTSE100
(UK)
$-JPY
$-DEM
(Japanese Yen) (Deutschemark)
NIK225
(Japan)
DAX30
(Germany)
HSI33
(Hong Kong)
GBP-$
(Sterling)
$-CAD
(Canadian $)
AUD-$
(Aussie $)
It is most commonly used to generate return in a rangebound market whereby buy low sell high strategies pay off.
The formula of slow stochastics (%D) used is as follows:
%D = (%Kt, %Kt-1, ...., %Kt-m+1) / m where m = smoothing factor
%Kt = (C-L) / (H-L) x 100
where C = closing price at time t.
Again, different parameter inputs are used to simulate traders/investors of different time-frame orientation. The full
range of parameters tested is as follows:
n = observation period
MTA JOURNAL
(10 days
;
5 days)
(30 days
;
5 days)
(30 days
;
10 days)
(60 days
;
5 days)
(60 days
;
10 days)
(60 days
;
30 days)
(100 days
;
10 days)
(100 days
;
30 days)
(100 days
;
60 days)
A buy signal is generated when the value of the stochastics
indicator drops below the predefined oversold threshold, and
vice versa.
In the test, we adopt the commonly used overbought and
oversold thresholds of 85% & 15% respectively.
Once stopped out of a position, a neutral position will be
maintained until a new signal is derived.
6) Defining the profitability of a trading system and the value added
by applying stop-loss:
a) A trading system/strategy is considered to be profitable if its
cumulative P&L, with NO stop-loss applied, is positive, and vice
versa.
b) Application of stop-loss is deemed to add value if it contributes positively to the cumulative P/L of the trading system,
and vice versa.
7) Frequency (probability) versus magnitude of the impact of stoploss:
In this study, we focus on the frequency/probability, rather than
the magnitude, of positive versus negative impacts of stop-loss. Owing to the diversity in the nature and characteristics of the markets tested, aggregating the magnitude of the P&L impacts is likely
to produce bias or unrepresentative results, in the sense that the
most volatile markets would dominate the test results. In order to
give the results from each market an equal weighting, we have
opted to focus on the frequency (i.e. % probability) rather than
the magnitude of impact.
SUMMARY OF RESULTS
Results of the test are shown in the appendix. Tables A to H summarize the results of individual tests. More important are tables 1a to
7b, which contain the key findings. These include:
1) The impact of stop-loss on the "pattern" versus the "probability"
of loss
The use of stop-loss changes the "pattern" of loss by reducing
the size of individual losses, but at the same time it increases the
frequency of losses. The net impact is uncertain and the overall
results are very mixed.
There is no significant conclusion we can draw from the crude
Spring-Summer 2001
35
2)
3)
4)
5)
results of individual tests2 which did not show any consistent impact of stop-loss on the final P&L.
However, a much clearer pattern emerges once we bisect the
test results according to the style and profitability of the trading
strategies tested (Tables 1a to 7b).
Impact of stop-loss on different trading styles (Tables 1a, 1b)
The application of stop-loss to trend-following trades generated mixed results. There is no significant evidence that it either
improves or undermines performance in a consistent manner (see
Table 1a).
The results from the application of stop-loss on mean-reversion trades are much more decisive. Stop-loss consistently improved
the performance of the mean-reversion trading strategies tested,
with a 73% probability. This result is consistent across different
markets, parameter values and stop-loss thresholds (see Table 1b).
Impact of stop-loss on profitable trading strategies (Tables 4a, 3a,
2a)
Stop-loss undermined the performance of the profitable trading systems with a 64% probability.
In only 31% of the cases, stop-loss further enhanced the results of the profitable trading strategies tested.
This observation applies to both mean-reversion and trendfollowing trading styles.
Impact of stop-loss on unprofitable trading strategies (Tables 4b,
3b, 2b)
Stop-loss improved the performance of unprofitable trading
strategies with a 73% probability.
In only 22% of the cases, stop-loss further exacerbated the loss
of unprofitable trading strategies.
This observation applies to both mean-reversion and trendfollowing trading styles. The results are very robust in both cases,
with a 84% and 62% probability of return improvement respectively.
Relative efficacy of trailing stop-loss versus fixed % stop-loss
(Tables 7a, 7b, 6a, 6b, 5a, 5b)
The results are not robust enough for us to draw conclusions
about the relative efficiency of the two types of stop-loss rules tested
(see Tables 7a & 7b).
However, there is a clear tendency for trailing stop-loss to outperform in the cases where the underlying trading strategies are
unprofitable and underperform in the cases where the underlying trading strategies are profitable. This can readily be seen by
comparing table 6a to 5a and table 6b to 5b.
QUALITATIVE ATTRIBUTES OF
STOP-LOSS MECHANISMS
Advantages of applying stop-losses
1) Reduce risk of ruin
The adoption of a stop-loss rule requires one to determine, in
advance, one's maximum "loss-tolerance" in a trade/position. In
this sense, the discipline of stop-loss will help to reduce the "risk
of ruin" regardless of its eventual net impact on the overall profit
and loss 3.
2) Avoid last minute decisions
Having a stop-loss discipline also helps to avoid the need to make
last minute decisions, which tend to be less well-thought out or
even irrational. When forced to make a last minute decision under pressure, investors are more likely to suffer from the "fear of
regret" syndrome. That is to say, having lost money in a position,
investors would hesitate to cut the position, fearing that the market price could reverse course, causing regret. Such "fear of regret" often works against one's investment process and discipline
and could significantly exacerbate the loss. A discipline in stoploss reduces the need to make last minute decisions and thus reduces the undesirable impact of "fear of regret."
MTA JOURNAL
Spring-Summer 2001
36
ENDNOTES
1
Note that the choice of markets to be tested is arbitrary and reflects the background of the author who specializes in equity and
currency market investment. The markets included in the test represent over 80% of global equity markets in terms of market capitalization and over 90% of FX trades in the global currency market on an average day. The test is readily replicable for investors
specializing in bond or commodity market investment.
Crude results are not shown in the appendix because the number
of data points are very large, and we are not drawing any conclusion based on the crude results. All our conclusions are based on
the summarized/bisected results shown on Tables A to H and table
1a to 7b.
For readers interested in the subject matter, the "Risk of Ruin"
article from Kaufman on Market Analysis, April 1994 provides a
insightful quantitative analysis.
BIBLIOGRAPHY
CONCLUSION
In this paper, we have studied, via both quantitative and qualitative analysis, the efficacy of stop-loss mechanisms. We have back-tested
two of the most commonly applied stop-loss rules (fixed % and trailing % stop-loss) on both trend-following and mean-reversion trading
strategies. The results from our tests indicate a significant negative
correlation between the value added by applying these stop-loss rules
and the profitability of the trading strategies tested. In other words,
there was a tendency for these stop-loss mechanisms to undermine
the performance of the profitable trading strategies tested and improve that of the unprofitable ones in the test. This observation is
true of both trend-following and mean-reversion trading strategies,
although the results are more robust statistically in the latter case.
The results lead us to query the expectation that stop-loss would
enhance trading performance over time by reducing the size of losses.
There are cases, certainly not a negligible minority as far as our test
results are concerned, whereby the tested stop-loss behaves in a fashion similar to an insurance policy, i.e. it generates a "loss" when the
underlying trading strategy works well, and vice versa. We therefore
argue against the blind acceptance of the utility of stop-loss mechanisms. It is advisable that investors should pre-examine (e.g. via backtesting) the impact of stop-loss mechanisms on their trading strategies/style before adopting them.
Besides, based on the test results, we also argue that a consistent
value added from stop-loss should not lead to complacency. Instead,
it should be taken as a warning signal that the trading strategy may
contain a particular weakness which is responsible for the consistent
profitability of the stop-losses. To the extent that is possible, such
underlying weakness should be overhauled directly rather than controlled by a stop-loss.
There are other advantages and disadvantages of applying stoploss which are not readily quantifiable. Namely, applying stop-loss
reduces the risk of ruin and avoids the need to make last minute
decisions. However, in applying stop-loss, investors would also suffer
from the difficult dilemma of re-entry. As each investor would attach
different importance to these qualities, it is difficult to draw any gen-
MTA JOURNAL
BIOGRAPHY
William K.N. Chan is Head of FX Strategy in HSBC Asset Management Limited. Before taking up his current position in London, he was a member of the Tactical Investment Unit of the
same company responsible for technical analysis and FX decisions. He also produced quantitative analysis on other areas of
investment management. One of his research articles on portfolio rebalancing was published in the Pensions & Investment
magazine.
William holds a M.Sc. degree in Economics from the L.S.E.,
University of London and a BSSc. degree in Business Studies
and Economies from the University of Hong Kong. He has been
a CFA since 1997.
Spring-Summer 2001
37
MTA JOURNAL
Spring-Summer 2001
38
MTA JOURNAL
Spring-Summer 2001
39
SUMMARY
If markets are efficient, then they should be interrelated. As such, the price
action of interest rates should be correlated with that of interest-rate-sensitive
stocks. Additionally, the "breadth" of the indexes should describe the health of
the prevailing price trend in interest rates. First this study determines that of
several popular interest-rate-sensitive stock indexes, (Dow Jones Utility Average, S&P Electric Company Index, The Philadelphia Bank Stock and Utility
Indexes, The Morgan Stanley Cyclical Index, the S&P Line Insurance Index,
and the S&P Bank Stock Index) only the Dow Utility Average, the Philadelphia Utility Index and S&P Electric Company Index from July 1977 to April
1997 are significantly correlated with bond prices (as defined by 100 minus
the bond yield). The study then uses subjective charting techniques and nonsubjective computer trading systems with the rules from the Martin Zweig's
Zweig Indicator and Gerald Appel's STIX Indicator to illustrate that in fact
the breadth of movement in the S&P Electric Company Index contains some
valuable information (forward-looking) for determining the health of bond
market trends.
MTA JOURNAL
Table I
Number of
Declining
stocks
Difference
(A-D)
Day
Number of
Advancing
stocks
A-D line
(Cumulative
running total)
200
300
-100
-100
400
100
300
200
300
200
100
300
450
50
400
700
100
400
-300
400
The ratio or simple difference calculation by itself acts as an oscillator with overbought and oversold levels. Like Col. Ayers, analysts
attempt to identify the extremes as an indication of buying climaxes
or selling capitulation. Contrarians generally use such evidence as a
signal that the market has exhausted itself and initiate counter-trend
trades.
All of the indicators and research presented in this section were
developed for use in the stock market. These indicators use NYSE
composite or other composite data in their development and application. Simple A-D indicators include:
Spring-Summer 2001
40
MTA JOURNAL
Spring-Summer 2001
41
of the expected values, thus bearish (the generals are leading the
army). A negative value is bullish, since the expected values are
pulling ahead of the DJIA (the army is leading the generals). For
trading signals, Merrill uses a 2/3rds Standard Deviation above
and below the mean as extremes.
MTA JOURNAL
Spring-Summer 2001
42
Methodology
As discussed, the A-D line is an effective tool because it captures
sector rotation. Investors sell weak stocks because they feel they are
vulnerable to an economic downturn. An extension of this theory
would logically be that if stock market participants believe that interest rates are going to rise, then they would sell stocks that are vulnerable to changes in interest rates.
Popular interest-rate-sensitive stock indexes are examined in this
study. First, a determination is made if these indexes are correlated
with interest rates. Then breadth indicators are constructed using
the price action of the stocks underlying the indexes. Using the traditional package of A-D line analysis (analysis of the line itself, moving averages, momentum and oscillators) the breadths of the interest rate stock indexes are tested to determine if they have any meaningful information concerning trends in interest rates. Daily data
from July 1979 to April 1997 were used in this study.
Interest Rates
Although there are several proxies for interest rates, the yield on
the U.S. Treasury 30-year bond is the most preferable for several reasons. Other possible proxies include The Dow Jones Bond Average
and the 30-year U.S. Treasury Bond futures contract traded at the
Chicago Board of Trade.
The Dow Jones Bond Average uses 20 bonds traded on the New
York Stock Exchange. The problem with this index is it contains
convertible bonds that can distort the actual movement of interest
rates. Companies subject to merger and acquisitions rumors that
have bonds with convertible options will respond to factors that have
nothing to do with interest rates, but perhaps the convertible options
being reevaluated. Additionally, changes in perceptions of credit
quality will affect this index even if expectations of interest rates don't
change.
The 30-year U.S. Treasury Bond futures contract traded at the
Chicago Board of Trade has two problems. The first problem is that
although it does trade off the underlying behavior of the actual 30year U.S. Treasury Bond, it is a futures contract, and thus by definition trades off a future expected outcome. The convergence doesn't
occur until the final month of the contract. The second problem is
the issue of the roll. The bond contract is a quarterly contract. Each
quarter, gaps in price occur due to one contract rolling off and another starting. These gaps are more pronounced than the "rolls" that
occur from the Treasury auctioning other bonds.
The proxy we have selected is the yield of the current U.S. Treasury 30-Year Bond. This probably is the most accurate proxy for interest rates because most analysts tend to look at the bond for trend
changes that could forewarn of changes in the equity market. Since
classical chart analysis is based upon price movement, the Treasury
Bond yield is converted into a "discount" price, 100-yield (yield and
price move in opposite directions).
Chart I
Philadelphia Utility Index and the Bond
Interest-Rate-Sensitive Indexes
It is generally thought that utilities, banks, life insurance companies and cyclical stocks are greatly influenced by interest rates. As
such the following eight indexes were analyzed (see Appendix I for a
list of the stocks that compose each index):
The Dow Jones Utility Average, an average of 15 utility companies
listed on the New York Stock Exchange that are involved in the
production of electrical energy and natural gas;
The Philadelphia Utilities Index, a capitalization-weighted index
MTA JOURNAL
Spring-Summer 2001
43
Chart III
Dow Jones Utility Average and the Bond
Chart I confirms the results of the correlation matrix: the Philadelphia Utility Index and the bond discount-price tend to move in
tandem. Upon closer examination, however, it appears that the Philadelphia Utility Index tends to bottom and peak before the bond discount-price. Chart II shows the S&P Electric Company Index and
the bond discount-price. Again, the S&P Electric Company index also
appears to bottom and peak before interest rates. Chart III shows the
Dow Jones Utility Average and the bond discount-price. Although
both series tend to move in tandem, the Dow Jones Utility Average fit
is not as good as either the Philadelphia Utility or S&P Electric Company index as the correlation matrix suggests.
(right scale)
Chart II
S&P Electric Index and the Bond
(left scale)
Bond is the "Discount" price defined as (100 - 30 year U.S. Treasury Bond yield)
(right scale)
Chart IV
Philadelphia Utility Index A/D and S&P Electric Company Index A/D
Weekly
(left scale)
Chart V
Philadelphia Utility Index A/D and S&P Electric Company Index A/D
Daily
Correlation of the Philadelphia Utility and S&P Electric Co. Index's A-D lines
7/79 4/97
1/79-12/94
1/96-12/96
1/97-4/97
99.2%
99.7%
88.0%
78.0%
The Philadelphia Utility Index A-D line and S&P Electric A-D line
are significantly correlated from July 1979 to December 1994. The
relationship, however, diverges beginning in 1996 suggesting that the
price action in the nine companies included in the S&P Electric Company Index has a material effect on the A-D line and further examination is needed.
Charts IV and V further illustrate this point that although the A-D
lines generated from the two indexes are similar in nature, there are
some striking differences. The period from late 1996 to April 1997 is
one such period. The A-D line in the S&P Electric Company Index
appears to continue its steep descent, while the Philadelphia Utility
Index A-D line appears to be leveling off.
MTA JOURNAL
Spring-Summer 2001
44
Chart VIII
Chart IX
Chart VII
Chart X
MTA JOURNAL
Spring-Summer 2001
45
Chart XI
MTA JOURNAL
Spring-Summer 2001
46
MTA JOURNAL
Spring-Summer 2001
47
CONCLUSIONS
It appears that the price action of interest-rate-sensitive stocks provide clues to trend changes in interest rates, as defined by the bond
discount-price (100-bond yield). The S&P Electric Company and
Philadelphia Utility Indexes are significantly correlated with bond
prices. Under subjective chart analysis, the indexes appear to lead
bond prices at turning points. Further analysis showed that the S&P
Electric Company Index held a stronger correlation to bond prices
across various periods and subsequently was used for the rest of the
analysis.
Using subjective chart analysis, an A-D line generated from the
price action of stocks in the S&P Electric Company index appears to
contain some information concerning the viability of bond price
trends from 1990-97. Although chart analysis on the A-D line prior to
1990 suggests that there is correlation between bond prices and the
indexes, analysis using a set of rules applied to a computer trading
system disputes that finding. The fact that the computer finds that
there exists some information that the subjective chart analysis didn't
find highlights the problem with the A-D line chart analysis: it can be
quite subjective both biased by analysts' preconceptions and inability to spot information based upon the presentation of the chart.
The computer, employing rules in a non-subjective manner, buys
or sells bonds depending upon the S&P Electric Company A-D ratio.
The Zweig Breadth and STIX indicators both contain significant information about trends in bond prices. The Zweig Breadth Indicator was not successful across all time periods tested and the large
intra-day drawdowns and inability to cut losses short makes the indicator/system a poor trading system to use. The Zweig indicator can
be use used as short-term indicator, but confirming indicators should
also be employed prior to trading off the signals. The STIX indicator, however, offered significant results across all time periods indicating it would be an excellent intermediate-term indicator for trend
changes in bond prices.
STIX
The STIX indicator is similar to the Zweig Breadth Indicator, except it employs a longer (21-day) exponential moving average. Gerald
Appel, the creator of the indicator, identifies 0.42 and 0.44 as oversold and 0.56 to 0.58 as overbought for NYSE composite stock data.
Across all time frames, the STIX system using the S&P Electric
Company A-D data for signals to buy and sell bonds is profitable.
With the exception of the 1994-95 period, over 50 percent of the
trades in each time period were profitable, generating a significant
net profit and return on equity. Although during the 1994-95 period, less than 50 percent of the trades were profitable, the system
was able to cut the losses short, so net profit for the period was positive. Again the computer is always in the market (long or short), but
the difference with the STIX rules is that both long and short trades
are profitable. The system was also profitable during the 1979-1990
time period, the same period where the subjective analysis deemed
the A-D line and trends in interest rates to be uncorrelated. The STIX
system appears to be a more balanced system. Thus the S&P Electric
Company Index A-D using the STIX rules is a good intermediateterm indicator for bond trends.
REFERENCES
McClellan Oscillator
The McClellan Oscillator is a short- to intermediate-term breadth
indicator based upon the same concept as Gerald Appel's Moving
Average Convergence-Divergence Oscillator (MACD). Essentially it
is the difference between a 19-period and 39-period exponential
moving average of advances less declines (A-D). Typically, the
McClellan Oscillator reaches an oversold or overbought extreme prior
to a change in the trend of the stock market. An oversold market is
identified by oscillator readings between -70 and -100. Oscillator readings of +70 to +100 indicate an overbought market.
It was extremely surprising that the McClellan Oscillator did not
return any profitable results over the same time periods tested for
the Zweig and STIX indicators. Even when run through an optimizer,
the Oscillator was unable to return significant results.
MTA JOURNAL
Spring-Summer 2001
48
MTA JOURNAL
BIOGRAPHY
Gary Stone is a VP of Business and Content Development
and Director of Finance and Strategic Business Planning for
Multicast Media, a startup company building a broadband broadcast overlay to the Internet to deliver high-quality streaming
media to the desktop. Prior to Multicast Media, he spent seven
years as a trader at Paribas Corporation in the Fixed Income
Department on the financing, short-term arbitrage, and U.S.
Agency desks. Mr. Stone was also an assistant economist in the
Domestic Research Division and a trader/analyst in the OpenMarket Trading Division of the Federal Reserve Bank of New
York. Mr. Stone has an undergraduate degree in economics and
computer sciences:mathematics from the University of Rochester and received an MBA from the Stern School of Business at
New York University.
Spring-Summer 2001
49
PREAMBLE
MTA JOURNAL
Spring-Summer 2001
50
Chart 1
Vertical Chart, Daily
San Francisco Company
Chart 2
Point & Figure
San Francisco Company
MTA JOURNAL
Spring-Summer 2001
51
Side Bar #1
Wyckoff Buying Tests: Nine Classic Tests for Accumulation
Nine Buying Tests (applied to an average or a stock after a decline)*
Indication:
Determined From:
Figure Chart
Vertical or Figure
Vertical
Vertical or Figure
Vertical or Figure
Vertical Chart
Figure Chart
Adapted with modifications from Jack K. Hutson, Editor, Charting the Market: The
Wyckoff Method (Technical Analysis, Inc., Seattle, Washington, 1986), page 87
Side Bar #2
New Wyckoff Buying Tests Modified for Re-accumulation
Buying Test Three requires judging the volume on the rising and
falling price waves in the trading range. A visual inspection reveals
that by point #16 on the SF chart, volume was expanding on the rallies and shrinking on the declines. By the time point #16 was reached
on the vertical chart, SF would have passed the test: Activity bullish. Turning once again to the figure chart discloses that in the
vicinity of point #10 the downward sloping supply line (dashed line
SS) was broken. Thus around point #10, the Fourth Buying Test was
passed. These four foregoing tests, although necessary, were not sufficient evidence of accumulation, so the trader had to remain patient
until all of the buying tests clearly revealed that a base had been
formed and that the evidence had accumulated to prove that the line
of least resistance was decidedly upward.
The next two Wyckoff Tests are crucial to the definition of an
upward line of least resistance. Buying Test Five is higher lows (higher
supports) and Buying Test Six is higher highs (higher tops). The
vertical line or bar chart of SF showed higher price lows along the
gradient of points #14, #16, and #18. In a parallel fashion, a series of
rising price peaks appeared at points #12, #13, #15, and #17. At points
#17 and #18, the trader-analyst could clearly declare that the higher
highs and higher lows had been reached, and, therefore, Wyckoff
Buying Tests Five and Six had been passed.
Points #15 and #16, and then again #17 and #18 on the charts,
may also be viewed as Jumps and Backups, hence legitimate junc-
MTA JOURNAL
Spring-Summer 2001
52
taken together revealed a lack of supply being pressed upon the market. Moreover, the interception of the rising support line C-C indicated that a sufficient correction in time and price had taken place
(Test Six). However, it was not until the subsequent surpassing of the
resistance along the $31 level on June 11, on wide upside price movement and expanded volume, that SF satisfied several other Re-accumulation Tests, such as Test Number One Resistance Line Broken
and Test Number Four, Higher High (price). Then at point #23, a
pullback to a Higher Low was executed (Test Three) and a count
of the Figure chart along the 31 level would have projected upward
to $37-39. However, this count was insufficient to confirm the earlier
price target projections of $47-49 taken along the $22 level. Hence,
Re-accumulation Test Number Eight was not passed. Moreover, a trade
taken at 31 also would have fallen short of the 3-1 reward-to-risk minimum because a stop would need to have been placed 3 points away at
28, and the re-accumulation count was only 8 points. Thus, Test Number Nine also failed. Presumably a pattern analyst could have labeled
the consolidation from #19 to #20 a pennant (Test Seven).
FAILED TESTS
To illustrate the new list of modified Wyckoff Tests for Re-accumulation that appear in Side Bar #2, let us return to the case study of
the SF Company. After the base had been completed, the Wyckofforiented trader entered a long option position at point #18 on the SF
charts. The SF stock then moved up sharply from point #18 to point
#19, where it encountered enough supply to halt its advance, and so
SF entered a period of hesitation and sideways movement starting at
#19. This period of hesitation commenced with a buying climax
around point # 19, which would also have alerted the trader of the
possible onset of re-accumulation before resumption of the upward
trend or even possibly distribution leading to a reversal of trend. The
trader, who was actually operating in SF at the time of this case study,
recounted his upside figure chart objective to $49 and chose to wait
out is interruption in the trend.
At point #20 the trader observed a Spring situation and so presumably he could have ventured a long position around the $29 level
(see Active Trader magazine, August 2000, for Springs and Upthrusts). At this juncture he could have consulted side bar #2 for
the checklist of Re-accumulation Tests. At point #20, he could have
concluded that Re-accumulation Buy Tests Number Two and Number Three had been passed. At point #20 the volume had dried up
considerably and the downside price progress was minimal, which
CONCLUSION
* Examples of these and other tests for re-accumulation are available in the Wyckoff
literature. In Basic Lecture Number 12 of the SMI/Wyckoff course, for instance, the
narrator counsels the student to place resting buy orders at the 1/2 re-tracement level
in order to add positions during corrections in a bull market. Elsewhere in the Wyckoff literature the student is admonished to purchase when the price intercepts and
encounters support along an important upward slanting demand line.
MTA JOURNAL
When SF reached the $49 level, the trader exited his SF options
position. He judged that the relatively high volume occurring in the
price-objective zone around $49 was sufficient reason to exit. To make
Spring-Summer 2001
53
the case for exiting even more enticing, the general market index
had started to weaken and diverge from the higher price set by SF
around $49.
There were targets outstanding at $51-$57, but this Wyckoff-oriented trader elected to take his profits at $49 because that was the
maximum effect of the cause built during the re-accumulation stepping-stone-count along the $35 line (point #22 to point #31). He
reckoned that he would have to weather another sideways to down
correction/consolidation as further preparation for the final advance.
He further reckoned that the risk did not justify waiting to capture
the final 8 points available beyond $49. Of course, as we can see
retrospectively, he exited prematurely because SF promptly advanced
to $54. (Upon further reflection, this Wyckoff trader said that he
would do the same thing again because bulls make money, bears
make money, and pigs get slaughtered.)
The case study of the San Francisco Company (SF) demonstrated
how, with the help the Wyckoff Nine Classic Buying Tests, an option trader could have entered favorable reward-to-risk long positions
just as the line-of-least resistance became defined with the passage of
the Nine Classic Tests for accumulation and as the stock was leaving the base formation. This case study also demonstrated how an
option trader could have later employed a new set of the Nine Reaccumulation Tests to both roll his contracts forward and to add to
his position. The fulfillment of the stepping stone confirming count
nature of this re-accumulation consolidation gave the trader added
reason to hold on to his positions until his longer-term base targets
were being reached at $49. Furthermore, the stepping stone confirming count provided an additional compelling reason for him to
exit his long options on the burst of strength as SF reached the $49
level.
In general, the Wyckoff Nine Classic Buying Tests and the set of
Nine New Tests for Re-accumulation can help investors and traders
to advance forward in their quest to control risk, ride winners and
take home maximum profits.
MTA JOURNAL
* For readers who recall their high school physics lessons, the law of the cause and
effect can be likened to Hooke's Law of Elasticity. Hookes Law declares the agitations up and down build up energy, the cause (e.g. agitating a metal coat hanger back
and forth) and the resultant effect (bend the hanger out of shape) expends energy in
an exactly one-to-one proportion to the preceding energy built up.
Spring-Summer 2001
54
cal chart ought to be used for trend analysis; however determining the potential extent of the move is the special provence of the
figure chart, sometimes referred to as the cause and effect chart.
2. Procedure. The building blocks of the figure chart are box size,
intraday data, number of reversal points and full-unit crossing.
Most commonly the box size is one point. Hence, intraday price
action must meet or exceed the full price levels to trigger a figure
chart entr y.
Reversal points are normally one point or three-point. For the
one point figure chart, a very special consideration to keep in
mind under the Wyckoff figure chart procedure is the necessity
of having at least two entries in any column. Many software programs change columns when price changes direction, even if only
a single entry exists in a column. To compensate for this, the
analyst must shift prices to create a column with at least two entries in a column before price can move to the next column. Hence
a quick down, up, down of one point each would remain in a single
column.
For larger moves, the analyst has the option to either relying upon
the three-point reversal or an increase in the box size.
3. Perspective. The analyst can visualize horizontal counts as fitting
within a saucer appearing bottom and a dome looking top. The
first count line should be conservative, nearest the lows, and be
considered as the minimum possible. The next count line will
usually be within the trading range, broader, and considered the
likely objective. Finally, the pullback following the upside jump
or valid/breakout creates the widest count and the highest upside count, and is thus the least conservative measurement (this is
the last-point-of-support that follows after a more important signof-strength).
MTA JOURNAL
REFERENCES
BIOGRAPHY
Henry O. (Hank) Pruden, Ph.D., is Professor of Business and
is Executive Director of The Institute for Technical Market Analysis at Golden Gate University, San Francisco, CA, and he is also
Editor of the Market Technicians Association Journal. Hank can
be reached at hpruden@ggu.edu, phone 415/442-6583 and
www.hankpruden.com.
This article was reviewed, edited and approved by Mr. David
Upshaw, CFA, CMT, Associate Editor, MTA Journal.
Spring-Summer 2001
55