44 - 1994 winter

WINTER 1994-SPRING 1995 ISSUE 44

A PUBLICATION OF THE MARKET TECHNICIANS ASSOCIATION

ONE WORLD TRADE CENTER, SUITE 4447 l NEW YORK, NEW YORK, 10048 l (212) 912-0995

MARKET TEXHNICIANS ASSOCIATION JOURNAL

Issue 44 Winter 1994 - Spring 1995

Editor

Henry 0. Pruden, Ph.D. Golden Gate University

San Francisco, California

Associate Editor

George A. Schade, Jr., CMT Scottsdale, Arizona

Manuscript Reviewers

Connie Brown, CMT Don Dillistone, CFA, CMT Richard C. Orr, Ph.D.

Elliott Wave International Cormorant Bay Chronos Corporation

Gainesville, Georgia Winnepeg, Manitoba Lexington, Massachusetts

John A. Carder, CMT Charles l? Kirkpatrick, III, CMT Eugene E. Peroni, Jr.

Topline Graphics Kirkpatrick and Company, Inc. Janney Montgomery Scott, Inc.

Boulder, Colorado Exeter, New Hampshire Philadelphia, Pennsylvania

Ann E Cody Michael J. Moody, CMT David L. Upshaw, CFA, CMT

Invest Financial Corporation Dorsey, Wright and Associates Lake Quivira, Kansas

Tampa, Florida Beverly Hills, California

Robert I. Webb, Ph.D.

Associate Professor and Paul Tudor Jones II Research Fellow

McIntire School of Commerce

University of Virginia Charlottesville, Virginia

Printer

Tritech Services

New York, New York

Publisher

Market Technicians Association One World Trade Center, Suite 4447

New York, New York 10048

MTA JOURNAL / WINTER 1994 - SPRING 1995 1

MARKET TECHNICIANS ASSOCIATION, INC.

Member and Affiliate Information

ELIGIBILITY: MEMBERSHIP is available to those “whose professional efforts are spent prac-

ticing financial technical analysis that is either made available to the investing public or becomes a primary input into an active portfolio management process or for whom technical analysis is

the basis of their decision-making process.” Applicants for membership must be engaged in the above capacity for five years and must be sponsored by three MTA members familiar with the

applicant’s work.

AFFILIATE category is available to individuals who are interested in keeping abreast of the field of technical analysis, but who don’t fully meet the requirements for membership. Privileges are

noted below.

DUES: Dues for Members and Affiliates are $200.00 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 $50.00 with the endorsement of a professor.

APPLICATION FEES: Applicants for membership will be charged a one-time, non-refundable application fee of $25.00; no fee for affiliates.

Benefits of MTA

Invitation to MTA Educational Meetings

Receive Monthly MTA Newsletter

Receive MTA Journal

Use of MTA Library

Regular Members

Yes

Yes

Yes

Yes

Affiliates

Yes

Yes

Yes

Yes

Participate on Various Committees Yes

Colleague of IFTA Yes

Eligible to Chair a Committee Yes

Eligible to Vote Yes

Yes

Yes

No

No

Annual Subscription to the MTA Journal for non-members-$50.00 (minimum two issues).

Single Issue of MTA Journal (including back issues)-$20.00 each for members and affiliates, and $30.00 for non-members.

2 MTA JOURNAL/WINTER 1994 - SPRING 1995

STYLE SHEET FOR THE SUBMISSION OF ARTICLES

MTA Editorial Policy

The MARKET TECHNICIANS ASSOCIATION JOURNAL is published by the Market Technicians Associa- tion, One World Trade Center, Suite 4447, New York, NY 10048 to promote the investigation and analysis of price and volume activities of the world’s financial markets. The MTA Journal is distributed to individuals (both academic and practitioner) and libraries in the United States, Canada, Europe and several other countries. The Journal is copyrighted by the Market Technicians Associa- tion and registered with the Library of Congress. All rights are reserved.

Style for the JZTA Journal

All papers submitted to the MTA Journal are references should be put at the end of the requested to have the following items as pre- article. Submission on disk is encouraged by requisites to consideration for publication: arrangement.

1. Short (one paragraph) biographical presentation for inclusion at the end of the accepted article upon publication. Name and affiliation will be shown under the title.

4. Greek characters should be avoided in the text and in all formulae.

5. Two submission copies are necessary.

2. All charts should be provided in camera-ready form and be properly labeled for text reference.

3. Paper should be submitted double-spaced if typewritten, in completed form on 8% by 11 inch paper. If both sides are used, care should be taken to use sufficiently heavy paper to avoid reverse side images. Footnotes and

Manuscript of any style will be received and examined, but upon acceptance, they should be prepared in accordance with the above policies.

Mail your manuscripts to:

Dr. Henry Pruden PO. Box 1348 Ross, CA 94957

MTA JOURNAL/WINTER 1994-SPRING1995 3

MARKET TECHNICIANS ASSOCIATION

Board of Directors, 1995-96

Officers/Office Manager

President Vice-President/Long Range Philip Erlanger, CMT Paul Desmond Phil Erlanger Research Lowry’s Reports, Inc. PO. Box 2680 631 U.S. Highway 1, #305 Acton, MA 01720 No. Palm Beach, FL 33408 5081263-2536 4071842-3514

Vice-President/Seminar James Bianco, CMT Arbor Trading Group Inc. 1000 Hart Road, #260 Barrington, IL 60010 708/304-1511

Treasurer Andrea Neumann HSBC Futures, Inc. 140 Broadway, 17th Fl. New York, NY 10005 2121825-9302

Secretary Michael Moody, CMT Dorsey, Wright & Associates 3579 East Foothill Blvd., #250 Pasadena, CA 91107 9091626-9666

MTA Office Manager Shelley Lebeck Market Technicians Association, Inc. 1 World Trade Center, Suite 4447 New York, NY 10048 2121912-0995 FAX: 2121912-1064 MTA BBS: 2121912-1058

Committee Chairpersons

Accreditation Philip Roth, CMT Dean Witter Reynolds 2 World Trade Center, 63rd Fl. New York, NY 10048 2121392-3516

IFTA Liaison Kenneth Tower, CMT UST Securities Corp. 5 Vaughn Drive Princeton, NJ 085435209 6091734-7747

Placement Vincent Butkiewicz HSBC Futures Inc. 140 Broadway, 17th Fl. New York, NY 10005 212/825-5884

Computer John Bohinger, CFA, CMT Bollinger Capital Mgmt. PO. Box 3358 Manhattan Beach, CA 90266 3101798-8855

Journal Dr. Henry Pruden PO. Box 1348 Ross, CA 94957 4151442-6583 or 4151453-4704

Programs Walter Burke, CMT M C M Moneywatch 1 Chase Manhattan Plaza, 37th Fl. New York, NY 10005 2121908-4325

Education Dodge Dorland, CMT Landor Investment Mgmt. 103 East 75th Street, #4F/E New York, NY 10021 2121737-1254

Library Linda Raschke LBR Group 2 No. Country Lakes Drive Marlton, NJ 08053 6091753-7715

Public Relations Ralph Vince 279 North Street Chagrin Falls, OH 44022 2161247-0073

Ethics & Standards John Baron, Jr., CMT Janney Montgomery & Scott 39 Public Square, #206 Wilkes-Barre, PA 18701 7171823-0152

Membership Julia Bussie A. G. Edwards & Sons 141 West Jackson, #201-A Chicago, IL 60604 3121554-4280

FbgiOlM

Mark Scott The Volume Investor 3 Piedmont Center, #210 Atlanta, GA 30305 404./231-1207

Foundation James Stewart, Jr., CMT NatWest Financial Markets Group 10 Exchange Place, 22nd Fl. Jersey City, NJ 07302 201/547-2910

Newsletter Andrew Addison Addison Investment PO. Box 402 Franklin, MA 02038 5081528-8678

Past President Mike Epstein Sherwood Securities 1 Exchange Plaza, 21st Fl. New York, NY 10006 212/482-2454

4 MTA JOURNAL /WINTER 1994 - SPRING 1995

TABLE OF CONTENTS

Technical Versus Location, Fundamental Analysis: Location, A View from Academe . . . . . . . . 8 Location . . . . . . . . . . . . . . . . . . . . . . . . . . .26 Hamid B. Shomali, Ph.D. Professor Tamalyn ll Crutchfield The author Shomali argues that technical analysis as observes that candlesticks are best used a discipline is most suitable for trading. for early warning indicators as opposed In the short-run, markets are driven by to independent trading tools. Using data human emotions; on the other hand, from silver and T-Bond futures, Tamalyn longer-term trends are associated with V Crutchfield shows how it is pertinent economic fundamentals. Dr. Shomali to employ several traditional Western notes that with the rise in “behavioral techniques as “timing indicators” along finance”, technical market analysis is with candlestick patterns. gaining academic respectability.

Intermarket Sentiment: System Testing Using Sentiment in for Consistent One Market to Call Profitability . . . . . . . . . . . . . . . . . . . . .34 Prices in Another . . . . . . . . . . . . . 10 Muneer Al Hulaibi Consistent Annual

Timothy W Hayes This paper shows Profitability (CAP) is a testing method,

that effective indicators can be produced not a trading rule. The purpose of CAP

using not only a market’s own sentiment, is to identify trading rules which pro-

but also the sentiment in other markets vide consistent year-to-year profitability.

as well. Tim Hayes studies the relation- Through the application of moving aver-

ships between price and sentiment in ages to the British Pound, the CAP test-

several markets, including gold, bonds and ing method revealed those variables

stocks using Market Vane sentiment data. which performed consistently well every year over a period of years.

MTA JOURNAL/WINTER 1994 - SPRING 1995 5

The Klinger Volume Membership Oscillator (KVO): and Affiliate A Theoretical Model . . . . . . . . . 45 Information . . . . . . . . . . . . . . . . . . . . . . .2 Stephen J. Klinger To arrive at the “volume force” or the volume fueling prices toward higher or lower levels, Style Sheet for

the KVO quantifies the difference be- the Submission

tween the number of shares being of Articles . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 accumulated and distributed each day in a security. Stephen J. Klinger provides a formula for calculating an oscillator of MTA Officers “volume force” (KVO) together with the and Committee results of empirical tests. Chairpersons.. . . . . . . . . . . . . . . . . . . . . .4

Improving Returns While Controlling Risk: Integrating

Editor’s Commentary . . . . . . . . . . . . . . . . . . . . . . 7

Wyckoff’s Tools with CANSLIM Stocks . . . . . . . . . . . . . . 53 J C. Coppola III An extensive study using O’Neill’s CANSLIM stocks reveals that adding the Wyckoff method of technical analysis substantially improved timing signals and performance results. Accumulation and distribution phases are described from the perspective of a Composite Operator. Wyckoff’s approach to position-taking, loss limitation, holding and selling are seen as aids to controlling risk.

6 MTA JOURNAL / WINTER 1994 - SPRING 1995

Editor’s Commentary

CMT III: Requirements, Topics and Guidance by Henry 0. Pruden, Ph.D., Editor

The following commentary was inspired by the interaction that occurred during the CMT III Work- shop at the 20th Annual MTA Conference in Las Vegas, Nevada, May H-21,1995. Approximately two dozen candidates for the CMT III met with the Editor and other members of the Accreditation Committee and the Journal staff to exchange views regarding the requirements, topics and guidance for the CMT III paper.

The reader can vicariously participate in the spirit and substance of the workshop by first reading and then reflecting upon the brief case study appearing immediately below:

Case: Del Hawkins Con@onts CMT III

Anxiety overcame Del Hawkins as he left the CMT II exam. He was apprehensive over his ability to switch mental gears from preparing for an exam to writing the CMT Illpaper: Rumor had it that several compe- tent technicians had given up before ever submitting their CMT Illpapers. Allegedly they believed that they had to prepare original, “walk on water”papers, and that the MTA reviewers tore articles to shreds before accepting them. These fearful reports left Del puzzled and anxious over whether he should make a “safe” study of Japanese Candlesticks or write something novel which interested him: the market as a mean- dering river

What should Del Hawkins write on? Why?

The requirements of the CMT III paper are chal- lenging and professional, but the candidate need not write a “walk on water” paper. The opening para- graphs of the guideline for preparing the CMT III paper are apparently forbidding. As a consequence of the workshop at the Seminar, it is the intention of the MTA Journal Editor and staff to make the intro- ductory guidelines more user friendly These revisions will be executed in concert with the Chair of the Accreditation Committee.

All participants at the workshop were encouraged to talk about their CMT paper topics. From the ensuing discussion it became clear that a modest, incremental contribution to the field of technical

analysis would be satisfactory. A number of promising topics were shared, including a test of daily candlesticks, a statistical technique, combining indicators into a pattern for decisions, sector studies, a look at underlying support and resistance, an investigation of indicator break down, point and figure analysis and so on. Almost invariably the promising CMT III topic was a subject which the candidate used regu- larly in his/her work.

In order to ease the transition from the more familiar method of examination to the less familiar enterprise of research and writing, candidates were strongly urged to: 1) start their CMT III projects upon completion of CMT I, and not to wait until after CMT II; 2) avail themselves of the mentors available through the MTA Accreditation Committee; 3) sub- mit an outline of their topic in order to gain valuable critical feedback and to be assigned a mentor. This more protracted, incremental process should ease the pain/anxiety candidates feel when confronting the CMT III.

Is the ultimate reward worth the pain and the effort of writing the paper for CMT III? The answer is a resounding yes. By completing a quality study, the candidate will have a concrete item of professional caliber to demonstrate to colleagues, clients or employers. Also a quality article raises the overall stature of technical analysis, which raises the professional standing of all CMT’s. Finally, the paper does lead to the coveted Chartered Market Techni- cians designation. The CMT designation alone is sufficient justification.

MTA JOURNAL/WINTER 1994 - SPRING 1995 7

Technical Versus Fundamental Analysis: A view from Academe by Hamid B. Shomali, Ph.D.

In the last several decades, there have been two com- of thought, or the fundamental analysts, regard peting schools of thought regarding the analysis and the valuation of financial assets determined by a valuation of financial securities. The traditional fi- “random walk”. In random walk, the prices of secu- nance experts have espoused fundamental financial rities are likened to steps of a drunken sailor, where analysis, dealing with identification of variables each step is independent of the previous one. Fun- which will determine the underlying value of securi- damental analysts basically assume an “efficient ties. These traditional security analysts have market” when the stock prices reflect all informa- downplayed the significance and the relevance of the tion available to the public. The efficient market other school, namely “Technical Analysis”. The strict theory, that the fundamentalists adhere to, assumes fundamentalists have viewed technical analysts as rationality at all times on the part of investors and mere “chartists” who pass over past data in order to does not allow behavior based on emotion and all find certain patterns in the behavior of security prices other impulses. Yet as a practical matter, human over time. These traditional finance theorists at irrationality is important. Even in the legal code, the times have likened the technical analysts to “Astrolo- plea of insanity allows for impetuous behavior that gers” in the field of finance. The technical analysts, is not based on rationality and simply stems from a on the other hand, would like to gain recognition for sudden urge or instant decision. Perhaps some of the their successes in forecasting security prices and be most persuasive evidence against the “efficient mar- considered more like astronomers than astrologers. ket” theory comes from the “anomaly” literature, But the debate continues. which has discovered unusual patterns in the price

In recent years technical analysis has been gain- behavior of securities. Some of the most puzzling ing wider acceptance in academia. Technical analy- price anomalies are related to seasonal patterns in sis received prominent and favorable review in a the movement of stock prices. Other anomalies relate seminal article surveying the frontiers of finance to returns that are dependent upon the size of a firm which appeared in the October 21,1993 issue of The and the impact of new stock issues. Economist. The offering of courses in technical analy- It is the contention of this author that in the sis at some universities such as Dartmouth College, short run (anything from a day to a few months), Golden Gate University and the McIntire School at emotions and other biases may lead us to make a the University of Virginia, as well as the New York decision that may not be based on rationality Im- Institute of Finance, demonstrates the increasing pulsive behavior, herd mentality or any other deci- recognition of the field of technical analysis by sion making process which relies on mechanisms academia. As technical analysts align their field with other than rational analysis of all relevant factors “behavioral finance”, they will gain even wider are not allowed in the fundamental analysis or the acceptance. theory of efficient markets. But how else can one

It is interesting to observe that both schools can explain the events such as markets behaving differ- be viable by explaining different behavior patterns ently on Monday mornings than Friday afternoons, at different time frames, and as such do not have to or that every year there is a sense of nervousness in be necessarily competing schools. The fundamental the markets around October? analysis rests on the assumption of a rational person One of the major problems that behavioral ana- who incorporates all of the relevant data concerning lysts have to face is that in their analysis of the a certain asset before making a decision about its market they often ignore the concept of probabil- acquisition. In that regard, the past history of the ity. In other words, they often sound as if they are asset is totally irrelevant. In other words, with all stating their forecasts with certainty. As a con- the past glory of IBM, if the fundamentals are point- sumer I may react to a 50% discount offer based ing towards a dismal outlook, the investor will dis- on sudden impulse, but such impulse may not dic- regard the past history In a sense, the rational school tate my actions every time I encounter such a


discount. The behavioral analysts have to specify that their technique is only for short-run decisions, and as such may be more useful to traders than institutional investors, such as pension plans, who are concerned with the long term returns on assets. The technical analysts also have to find a way to incorporate probability analysis into their analysis. Other- wise, there is no basic problem with their use of past data to arrive at certain conclusions about the future. In traditional forecasting models, such as “time series analysis” such as “Box-Jenkins” the past data is also used to make inferences about the future. In fact in econometric forecasting, the “least square estimation” or “maximum likelihood” method, the forecast of a dependent variable is based on a weighted average of the past observations of the same variable. The above statistical methods simply determine the weights through statistical manipulation, and the forecasts are based on probabilistic assump- tions about the behavior of variables, and as such are not deterministic numbers. In fact, “autore- gressive” estimation methods are an important part of econometrics where the past values of a variable are used to determine its future forecast.

The fundamental analysts can point to the strength of an underlying security based on the fundamental variables which will impact its value in the future. But this analysis, by its nature, is a long-term phenomenon which is incapable of pin- pointing the time that such movement will begin. In other words, the fundamental analysts can never provide us with the turning point.

Should we dismiss one theory in favor of another? The study of these two methods of security analysis reveal that they are concerned with different time horizons and different decision making processes. Fundamental variables can certainly affect the value of a security over the long-run. However, in the last few years, the economists have accepted that there are a lot of human emotions entering the process of decision making, not just calculating rational behavior, at least in the short-run. Granting of the Nobel Prize in economics in 1993 to Professor Douglas North is testimony of admission by main- stay economists that other modes of behavior such as culture, habit, bias and prejudice as well as im- pulsive or random behavior could be used to explain consumer behavior. Increased attention paid to “behavioral finance” by some well known finance scholars should open the door for a less biased approach toward “Technical Analysis” by the traditional finance professors. The fact that industry, such as Japan’s, has decided to invest $30 million in researching such topics indicates the security industry’s serious interest in the topic of technical analysis.

Hamid Shomali, Ph.D., is professor of Finance and Economics, and Dean of the School of Business at Golden Gate University. Dean Shomali joined Golden Gate University in 1986 after a distinguished career in banking and finance. At the Bank of America, he completed several policy studies which impacted the international lending of the bank. Also as a member of the energy lendinggroup, he made a substantial contribution to the bank’s energy loan portfolio. As Deputy Managing Director of Bank Farhangian, Iran, he managed the bank’s construction and mortgage lending as well as its international operations. Prior to that he was an economist for the Central Bank ofIran where he completed analytical projects on a broad range of macroeconomic and monetary issues. Dean Shomali has served on the faculty of several universities including the University of California at Berkeley, University of Houston and the National University of Iran. His teaching and research has been in international trade and finance as well as oil economics. Dean Shomali consults with international companies on banking, finance and international management. Dean Shomali received a Ph.D. in Economics from the University of California in Los Angeles (UCLA) in 1973. His under- graduate education was completed at the University of Salford, England where he received a B.S. degree in Mathematics and Economics (Joint Honors) in 1968.

MTA JOURNAL /WINTER 1994 - SPRING 1995 9

r

Intermarket Sentiment: Using Sentiment in One Market to Call Prices in Another by Timothy W. Hayes

When using technical analysis to analyze a market, analysis in that regard-sentiment indicators can a good starting point is price action and volume. also often identify peaks and troughs at about the Overbought/oversold and sentiment analysis would time they occur, but they carry the risk of generat- usually come next. And at some point a thorough ing an ill-timed signal. The primary tenet behind analytical assessment would need to incorporate sentiment indicators is that “the crowd is always intermarket analysis-that is, how one market is wrong at extremes.” As Ned Davis explains in Being affected by the technical condition of another mar- Right or Making Money, “almost by definition, a top ket. This is generally done through an assessment in the market is the point of maximum optimism of the other market’s price action. But can another and a bottom in the market is the point of maximum market’s sentiment conditions, in and of themselves, pessimism.“2 The tricky part is determining when serve as useful market indicators? This paper will those points have been reached, as no single indica- explore that question. tor can assess the entire psychology within a mar-

Before getting started, however, let’s look sepa- ket. Sentiment indicators can often be early at major rately at intermarket analysis and usual technical tops, or sometimes late. And since indicators will analysis, and their various pros and cons. Intermurket rarely reverse from the exact same level, identifying anulysis is valuable for its ability to identify condi- the extremes can be next to impossible except in hind- tions in one market that have been consistent with sight. This is why before changing a position, it is bullish or bearish performance in another market. usually a good idea to wait for the sentiment indica- In 1987, for instance, commodity prices rose, the U.S. tor to first identify an extreme, and to then confirm dollar weakened and bonds fell, each issuing a warn- that extreme by indicating a sentiment move in the ing for the stock market prior to the October crash. opposite direction. The benefits of intermarket analysis were well docu- One of the primary benefits of sentiment indica- mented in John Murphy’s Intermarket Technical tors is that they can place tape action in a longer- Analysis.’ And indicators used by Ned Davis Research term context and help you determine the degree to also show how the price action of bonds, the CRB which upside potential is greater than downside risk, Index, and the U.S. dollar can generate effective and vice versa. A pure tape-action gauge may indi- signals on the stock market. cate that the market has entered a downtrend. The

The pitfall of using intermarket analysis, how- sentiment indicator can help you to determine if the ever, is that the “normal” relationship can at times downtrend is most likely a correction within a longer- break down. This is not true of indicators based term uptrend (i.e., the sentiment is far from extreme purely on the market’s own action. Whereas an optimism levels) or whether the decline is more likely indicator based on a market’s own 200-day moving the start of a longer-term downtrend (i.e., the senti- average is certain to eventually flash a signal during ment has reversed from extreme-optimism levels). a major, prolonged move, such a guarantee does not When used carefully, sentiment indicators can help exist with intermarket indicators. T-Bill yields, for you identify tops and bottoms within the time frame example, usually bottom prior to bull market peaks of your concern. and top prior to bear market bottoms. But from 1976 What is the best way to use sentiment indica- into 1960, the yields kept rising straight through a tors? The answer is probably to use as many reli- bear market, a complete bull market, and nearly all able ones as possible, on a composite basis. And of another bear market. At times, inter-market indica- this is where intermarket analysis comes in. With tors can offer the top and bottom spotting capability the increase in information about competing in- that a trend-following indicator could never provide, vestment vehicles and with their increased acces- though at other times they can be inaccurate, and sibility, the investor’s options have expanded enor- therefore costly mously-assets can quickly be switched not only

Sentiment analysis is similar to intermarket between equity mutual funds, but also between


markets themselves. So if, for example, a peak in optimism would be evident in Market A alongside peaks in pessimism in Markets B and C, it stands to reason that as newfound pessimism replaced waning optimism toward Market A, and as grow- ing optimism replaced receding pessimism toward markets B and C, funds would start shifting away from Market A and toward Markets B and C. This leads to the broader question of whether historical tendencies can be found for sentiment extremes in one market to indicate the direction for prices in another market.

The bigger challenge is to determine whether these sentiment data can be useful for indicating price reversals in other markets.

The Methodology Before drawing any conclusions, a thorough

analysis must be undertaken to determine whether each data series can in fact be used to generate effective buy and sell signals. And before undertaking the data analyses, a methodology must be developed. I therefore went to work with the computer to produce a consistent methodology for the testing. This included:

The Data For testing purposes, I’ve chosen three markets-

the stock market, the bond market and gold. Although sentiment data are widely available for all three markets, the sentiment data series chosen for the three markets must be comparable. One of the most popular types of sentiment indicators is the put/call ratio, and within the stock market alone, various types of ratios are available from various sources, such as Merrill Lynch and the Chicago Board of Options Ex- change. But comparable put/call data for gold and the bond market are lacking. Investors Intelligences has excellent data on advisory service sentiment, but their data are limited to the stock market. Consen- sus Inc.4 is another good source for sentiment data, and it is available for stocks, bonds, gold and numerous other markets. But the drawback is relatively limited historical data for testing purposes.

Market Vane5 is a data source that does provide the same type of sentiment data for numerous markets and does so with a relatively extensive data history Through its Bullish Consensus service, Mar- ket Vane surveys futures traders about their views toward the markets and then reports the overall results on a scale from O%, or extreme pessimism, to lOO%, or extreme optimism. Since the data are reported in the same manner for each market, it is a very suitable data source for comparing sentiment in different markets.

Of course, no sentiment data series is perfect, and the Market Vane data is no exception. when used without any manipulation, the Market Vane data for gold, Treasury Bond futures and stock index futures have had little if any correlation with their respective markets, as indicated by their respective correlation coefficients of .lO, -.lO, and -.15 over the 16- year period from l/05/79 though l/06/95. This may reflect the high volatility of the Market Vane data, especially prior to 1987. But when Market Vane data are manipulated by an optimally-determined moving average, the volatility is smoothed out and effective Market Vane sentiment indicators can be developed for the respective markets, as will be shown later.

* moving average (“smoothing”) optimization l slope analysis * channel analysis l volatility band analysis

The slope analysis looked at reversals in the sentiment data and indicated how far, in terms of percentage or points, the actual or smoothed sentiment data would have to move to generate the best signals. The channel analysis identified the best fixed sentiment level for buying and the best single level for selling. A decision rule of this analysis required that for a signal, the data would have to rise above the upper fixed level and then reverse back below that level, and for the opposite signal, the data would have to drop below the lower fixed level and rise back above that level.

The volatility band analysis went a step further, using the concept behind John Bollinger’s “Bollinger bands.“‘j This analysis worked the same way as the channel analysis, looking for the optimum buy and sell levels. But the big difference with bands is that instead of fixed levels, the levels fluctuate, determined by an optimum number of standard deviations above and below the smoothed data’s mean for an optimum number of weeks. When the data moves beyond these levels, it is an outlier in the data’s bell- shaped curve for the time period used.l

The bands for one of these indicators might, for example, be determined by a single standard deviation above and below the latest 20-week mean of the sentiment data’s 50-week moving average. And the indicator might then require that for a buy signal, the smoothed sentiment data must drop below the lower band and then rise back above it, and that for a sell signal, the smoothed data must rise above the upper band and then fall back below it. Such indicators are therefore the result of four variables-the moving average of the data, the moving mean from which the standard deviation levels are determined, the number of standard deviations, and the method of signal generation

J


(this analysis includes the determination of whether to buy high and sell low, or do the opposite).

Thus lending themselves to more optimization than other types of indicators, volatility band indicators can be fitted closely to catch major highs and lows. Such indicators should therefore be examined closely to make sure that the historical performance of an optimized indicator is not the result of a few fortunate encounters between the bands and the smoothed data. Moreover, they should be examined in an out-of-sample test period, as explained later. When analyzing with so many degrees of freedom, the potential for curve-fitting cannot be underesti- mated, and it must be addressed.

The big advantage of indicators that use these types of parameters is that they automatically adjust for shifts in the data’s volatility and in changes in the range between its extremes. This advantage is critical when using the Market Vane data, which became less volatile starting in 1987, when its extremes started to become closer. The bands move with the smoothed data, identifying the levels that should be considered extreme based on the data’s recent history

So using the various analytical methods, I tested each of the three markets using the Market Vane data for gold, bond futures, and stock futures. I sought the best indicators that use the market’s own sentiment, and the best indicators that use the sentiment of the other two markets-nine relationships in total.

As it turned out, volatility band indicators worked best for each relationship, doing so by requiring that the smoothed data move through a band, reach an extreme, and then reverse back through the same band. These indicators often produced the best hypothetical results as determined by such statistics as the accuracy rate-the percentage of trades that would have been closed out for a profit; and the gain per annum-the compounded annual return that the indicator would have generated.s

But the assessment of what “works best” was based on more than just the best track record following optimization. When using the computer to crunch through a multitude of possibilities, the danger is that a good result is an exception caused simply by enough variables falling in line to call enough market turns to produce a good result. This is why it’s necessary to make sure that an indicator’s results are supported by similar results from similar types of indicators. It is also why it’s important to inspect the indicators visually in chart form. But most importantly, it is why out-of-sample testing is necessary, along with other criteria confirming that the indicator can be expected to work on a real- time basis.

After the initial testing process revealed that reversals through volatility bands would produce the most effective indicators, I analyzed each relationship again over the eight-year period from the beginning of 1979 to the beginning of 1987-the first half of the data’s date range. The resulting indicators with the best hypothetical track records were then applied to an eight-year test period from the beginning of 1987 to the beginning of 1995. This would indicate whether the optimized indicators would hold up if actually put to use at the end of the optimization period, continuing to produce effective signals despite changes in the sentiment data’s volatility

Along with the real-time test, I applied “friction” to make the results realistic when considering commission costs and slippage. Over the optimization period as well as the test period, a 0.25% cost was applied to every trade, this representing a common denominator for the three markets tested. And to make sure that the real-time results had statistical significance, I required a minimum of 30 trades in the test period. In summary, then, each indicator had to prove useful.. .

1. over the eight-year test period following the eight-year optimization period;

2. after accounting for slippage and commission costs of 0.25% per trade;

3. with at least 30 trades in the test period.

Bwdr Gold 22 SMA 6 0.8 BUy sell

BOOtiS 14 Eh4A 13 0.9 sell BUy

staks asMA 9 IS sell Buy CompoJice - _.

I I I / I I

~Stocks 1 Gold 1 I3!ZMA 1 1 1 IS 1 Buy 1 Sell

lstocks I 2FMA 1 19 1 0.7 1 sell j Buy

I ’ ,

To maintain consistency for comparative purposes, I sought indicators that would become opera- tional in close proximity Whereas an indicator based on the two-week average of a two-week smoothing would be able to generate a signal after four weeks, an indicator based on the 50-week mean of a 75-week smoothing would have to wait 125 weeks before a signal would be possible, making comparison difficult. As shown in Table #1, which summarizes the

12 MTA JOURNALa /WINTER 1994 - SPRING 1995

indicator formulas, the selected indicators ranged from a total duration (smoothing plus mean) of 16 weeks to 28 weeks.

An additional issue when assessing the results is whether one could actually trade at the market’s price at the time of the signal. For an indicator that uses S&P 500 closing price data to generate signals on the S&P 500, for example, realistic results would need to account for the fact that a trader could not act upon the signal until the next day’s opening, which means that he would be buying or selling at a different price than the closing price. When using the Market Vane data, however, the use of closing price data is realistic since the sentiment data is available on the Fri- day morning prior to the weekly close. Any weekly indicator using the Market Vane data could therefore be updated during the day, and the trader could then place a buy or sell order to take effect with the closing price.

Let’s now look at the analysis and results for each market individually, starting in each case with the market’s own sentiment and then looking at the market’s relationship with the sentiment of the other two markets. In the case of each market, it will first be demonstrated how a market’s own sentiment can be used to produce effective signals, and it will then

I

Stab IGold 1 I ,

33 1 70 I 15.1 1 13.4 1 31 ( 68 1 5.7 1 6.8 I .%a I 44 I .n I 10.8 I 13.1 I 46 I 54 I 7.9 1 5.7

be shown how, and to what degree, intermarket sentiment indicators can be used to produce useful indicators. For each indicator, the real-time results are shown in chart form (see Explanatory Churt) and summarized in Table #2 along with the results for the optimization period.

-

-

493

469

446

424

404

384

366

348

331

315

-

- EXPLANATORY CHART

Market Plotted in Top Clip and Called With Signals oat0 Fleqwncy - mnbd shown - ScMJg

MMJSNJMMJSNJYYJSNJMYJSNJMMJSNJJdMJSNJMMJSNJMMJSN 19aa 19a9 1990 1991 1992 1993

I

Pmfitabla Tradas: Pwcantage of Trades Clwed Wfih a Profit After Transaction Costs Gain Per Annum: Compounded Annual % Return Produced by Going Long on Buys

(6 Arrows), Sefling Shun on Sells IS Arrows), Including Transaction Costs Buy-Hold GPA: Equivalent w Annum Return Roducad by Staying Long During the Signal-Date Period

Signal Dates: Pwiod Prom lndkator’s Pimt Signal to End of Test Period

i e

‘: Arrows - 8bnecOmwation Pohts (Arrows in Top Clip Show Same Pohts Against Market)

-

493

469

446

424

404

384

366

348

331

315

-

L

80

16

72

68

64

60

56

52

48

44

40

36

32 -

Market Vane Sentiment Index and Moving Average Plotted in Lower Clip


THE ANALYSIS sentiment could be used to effectively call the gold market. In contrast to indicators that use gold sen-

The Gold Market timent to call gold, with sell signals following opti- A number of good possibilities resulted from the mism peaks and buy signals following pessimism

analysis of thegoldprice versus gold sentiment. peaks, the best gold indicators using bond sentiment The concept is simple-sell after an extreme in opti- proved to be those that flash gold buy signals follow- mism is evident, and buy once an extreme in pessi- ing bond optimism peaks and gold sell signals mism is indicated. But historical analysis of the gold following bond pessimism peaks. This suggests that market since 1978 must be handled with care due to crowd euphoria about the bond market reflects a the huge advance from 1979 to 1980. When compar- widespread view that inflation is dead, the condition ing indicators with the best per annum gains since that is usually bearish for gold. From a contrarian then, they will invariably be on buy signals during standpoint, then, a bond optimism peak would be the 1979-1980 period, and that one signal will have generally consistent with a trough in inflation a significant influence on an indicator’s overall his- expectations, and thus the gold price as well. Accord- torical gain per annum. So if an indicator is to be ingly, a bond pessimism peak is generally consistent considered reliable, it must have a good record after with a peak in inflation expectations and thus the that period as well, proving to be effective in the gold price. eight-year test period. The relationship between bond sentiment and

The top indicator that emerged from the analy- gold can be seen in Chart #2, which shows a ma- sis is based on the seven-week simple moving aver- jor low in bond market sentiment occurring in late age of the sentiment data for gold. The indicator uses 1987 at about the same time as a major peak in brackets that are 1.3 standard deviations from the gold. As bond sentiment remained in an uptrend smoothing’s nine-week mean, generating signals until its peak in 1989, the gold price remained in a upon reversals above the lower band from below downtrend until a trough at about the same time. (buy) and below the upper bracket from above (sell). Gold subsequently broke to the upside as the bond The indicator had a 62% accuracy rate for 37 trades sentiment broke to the downside, and the inverse in the optimization period while producing gains at relationship generally held true through the next a rate of 35.4% per annum, more than seven times four major reversals. From 1991 into 1993, the the equivalent buy-hold return of LO%-and again, bond sentiment remained in a low neutral trading this accounts for commission costs and slippage. range with repeated resistance at around 50, while

As might be expected, the per annum gain was gold formed a bottom. boosted by a profit of more than 100% from the 1979- The inverse relationship can thus be seen when 1980 long position. So it was no surprise to see the examining the longer-term trends, but it wasn’t per annum return drop to 4.2% for the test period, easy to develop an indicator that would generate with the ill-timed sell signal of March 1993 proving effective shorter-term signals. The selected indi- to be the most costly But as shown in Chart #1 (see cator has tight brackets that are 1.3 standard number in chart’s lower left hand corner), that sig- deviations from the smoothing’s mean of just four nal was an exception, as many of the real-time signals weeks, which makes it susceptible to whipsaws. were quite timely-especially the buy signals of But the test-period results did produce several very March 1987 and September 1989, and the sell signal timely real-time signals, such as the mid-1990 sell of January 1988. Furthermore, the accuracy rate signal and the mid-1993 buy signal. On the other during the test period-59%-was almost as good as hand, it failed to flash a buy just after the 1989 the 62% accuracy rate for the optimization period. sentiment peak. On balance, the indicator was

The tricky aspect of volatility bands is that they mildly successful in the optimization period as well can at times just miss major turns in the sentiment as the test period, though it did manage to beat data. And at other times, faulty signals can be the buy-hold return for the test period, doing so produced after the bands have been tracking the data with an accuracy rate that was better than that of sharply upward or downward, enabling the indica- the optimization period (see Table #2). tor to produce a signal upon a brief encounter When using the stock market sentiment to between the data and the band, as exemplified by develop an indicator for gold, one might expect the the October 1992 buy signal. But the test-period effort to be even more difficult since the relation- results do argue that in most cases, gold sentiment ship between stock sentiment and gold would seem can be relied upon to generate accurate signals for to be less direct. Whereas the gold and bond mar- gold using this indicator. kets are both directly at&&d by inflation, it might

With the next round of analysis came the first be argued that the stock market is impacted indi- real test of intermarket sentiment-to see if bond rectly, as it is more responsive to the bond market’s


Bands = 1.3 Standard Dwiiions From 8-W&x Ma

(1) Market Vane Gold Sentiment -- 7-Week Simple MovingAverage -__ --

l Examples of Timely Signals

BMds=1.3stmdard DwMonsFrwn4-W&MM

OMovesAtdveUppaBmdaThalBabwn Moves Bdow Lowr Bmd & mm Above r


COMEX Gold Futures (Nearest Contract) Weekly Darn I/09/87 - l/06/95 (Log Scale)

501 -

482 -

465 -

448 -

432 -

416 -

401 -c

387 -

373 -

359 -

346 -

334 -

322 -

8

i i

fi

n B

Profitable Trades: 57% Gain Per Annum: 5.6% Buy-Hokl GPA: -3.0% SiOllBl DBtBB: 7/31187 - l/6/95

Bends = 0.6 Standard Deviatii From &We& Mean Buy Sits = Smoothmo .Moves /ibOVB Uppsr Band & Thm Below It Sell Sipnsls = Smoothha Moves Below Lower Sand & Thm Above it

c

501

- 482

- 465

- 448

- 432

416

S

4

387

373

359

346

334

322 I I

MMISNIHMISNJMM1SNIMMJSNJMMJSNJtdMJSNJUMJSNJMMJSNJ 1988 19u9 1990 1991 1992 1993 1994 199:

I L-

i 60

1 58

56

54

52

SO

1 48

46

44

42

40

38 -

l- 38

(3) Market Vane Stock Sentiment -- 16-Week Simple Moving Average

reaction-i.e., it is more directly affected by interest rates themselves.

But a comparison of the sentiment series in Chart #2 and Chart #3 reveals a similarity in the long-term trends, though the stock sentiment’s influence on the gold price is less obvious. As shown in the test period results, a useful indicator can be produced when, as with the bond sentiment, gold buy signals are flashed after peaks in stock market optimism and gold sell signals are flashed after peaks in stock market pessimism. Based on the sentiment data’s 16-week smoothing, the indicator looks unusual due to narrow bands caused by a standard deviation of just 0.6 around an eight-week mean. But the indicator is in fact successful at catching gold reversals, doing so with most of the signals occurring when they would be expected to-i.e., buy signals just after peaks in stock optimism and sell signals just after extremes in stock pessimism. This type of consistency was lacking with the gold indicator based on bond sentiment.

The effectiveness of the stock market indicator is underscored by its ability to hold up on a real- time basis. In the optimization period, its per annum return was 22.1%, more than three times the buy-

hold result. In the test period, the per annum gain dropped to 5.6%, but this compares with a buy-hold return that fell to -3.0%. And the signal accuracy rate in the test period was 57%, close to the optimization-period accuracy rate of 60%. Also noteworthy was the test period’s lower maximum drawdown- i.e., the maximum loss that could have possibly been realized from any signal had the position been closed prior to the subsequent signal. This statistic improved from -24.4% for the optimization period to -6.6% for the test period.g

In summary, then, the analysis showed that an effective gold indicator could be developed using sentiment data on the stock market.

The Bond Market Moving to the bond market, the sentiment data

for bond futures yielded many good possibilities to choose from after analysis in the optimization period, suggesting that the two series have a reliable relationship. The test period returns were less impres- sive, though the indicator shown on Chart #4 did reliably indicate turns in the sentiment data, in the process generating timely sell signals in early 1988, mid-1989, mid-1992, and late 1993, among others.


Treasury Bond Futures (13-Week Perpetual Contract) Weskly Dam l/09/67 - l/06/95 (Log Scale)

Profitable Tradea: 59% Gain Per Annum: 4.4% W-Hold GPA: -0.1% Signal Dates l/23/97 - l/06/95

*

+ Examples of Timely Signals

Bands = 0.9 9tandard Deviations From 13-W& Mean Buy S&IBIB = Smoothhg MOWS Bekw Lower Band & Then Above tt SeJI Sbnab = smOothin Moves Above Upper Band & Then Below It

- 85

- 81

66 - - 66

63 -8 - 63

- 60

57

- 54

- 51

- 48

(4) Market Vane Bond Sentiment -- 1CWeek Exponential Moving Average __-

Several of the buy signals were also effective, occurring soon after significant bottoms. In fact, after looking at all the signals, one might expect to see bigger returns, though the indicator did outpace buy-hold by a substantial margin with a clear majority (57%) of profitable signals. (Although the indicator’s 30th signal was still open at the end of the test period on l/06/95, it was closed-out with a profit in early 1995. It is therefore included even though only 29 trades had been closed out by l/06/95.) The bond indicator that uses bond sentiment data thus proved effective, showing that it’s usually a good idea to exit bonds after a peak in bond market optimism and to enter bonds after an extreme in bond market pessimism.

When using the sentiment data for gold to analyze the bond market, the results from the optimization period showed that in the same way that extremes in bond market optimism can be followed by effective buy signals for gold, with the opposite holding true for pessimism extremes, the extremes in gold optimism can be followed by effective buy signals for bonds, again with the opposite applying to pessimism extremes. Gold sentiment thus appears to have the same effect on the bond market that bond sentiment has on the gold market.

In the test period, however, the per annum gains declined substantially from the optimization period, with the selected indicator managing a marginal per annum profit. But its accuracy rate was slightly better in the test period, and the test-period trends evident in Chart #5 show changes in the trend of gold sentiment marking changes in the bond price trend, usually with a lead. From mid-1987 into 1989, gold sentiment trended downward while bond prices trended upward. After gold sentiment broke to the upside in late 1989, bonds started a correction that didn’t end until after gold sentiment had headed downward. When gold sentiment started rising again in early 1993, it foretold the major bond price peak and decline that would follow.

The results of the gold sentiment to bond price association were somewhat paradoxical. As one might expect, the signals corresponding to the major sentiment reversals were generally early, and the tight bands made the indicator prone to whipsaws. And the real-time per annum gain was only marginally profitable. Yet, the indicator had a 63% accuracy rate in the test period, and the chart does illustrate how gold sentiment can be used aa an indicator for bonds.

Like the gold sentiment analysis, the analysis of


51 -

48 - - 40

45 - - 45

42 - - 42

39 - Bond Weakness

(5) Market Vane Gold Sentiment -- 22-Week Simple Moving Avs ~____

Mid-l 999 Sell


Standard & Poor’s 500 Stock Index Weekly Dora l/09/87 - 1106195 (7.06 Scale)

8,

480 -

440 -

403 -

370 -

339 -

311 -

285 -

262 -

240 -

“WI. rw NIIUII. ,.cv”

Buy-Hold GPA: 6.7%

SInal Ostes 32Ol07 - 116195

l Examples of Timely Signals

Bmdo = 0.7 Standwd Owbtbno From 1 B-Week Ma

Buy S&tab = Smwthhg Moves Bebw Lower Eand & Then Abovs It

Sell S@ub = Smoothho Movea Above Upper Eand & Thm Bebw It

480

- 440

- 403

- 370

339

- 311

285

- 262

- 240

I _,,,,___..: .,.,....,_. ,,~..,,.,.,,....,....,.,,..........,....,......,..,........,..........1 I’

YMJSNJYYJSNJYMJSNJYMJSNJMMJSNJbfMJSNJMMJSNJMMJSNJ 8 1919 1992 1993 1994 1995

- 63

- 60

- 57

- 54

51

- 48

- 45

- 42

- 39

36

33

(71 Market Vane Stock Sentiment -- Two-Week Front-Welgntea MOVlng Average -___

stock market sentiment yielded many indicators with hefty returns in the optimization period. And these stock sentiment indicators held up better in the test period. Reflecting the tendency for the bond market to move in the same direction as the stock market, bond market sell signals followed extremes in stock market optimism and bond market buy signals occurred after extremes in stock market pessimism. Effective indicators could be found despite the tendency for the bond market to lead the stock market. The indicator shown on Chati #6 adjusts for the leading tendency by flashing many of its signals early-the indicator’s formula enables the data to at times move in close proximity to one of the bands, allowing the data to move through a band from one side and then back through it from the other side prior to the data’s peak, producing a number of signals before the actual data reversals. In 1991, for example, the bond market sell signal appeared in advance of the actual reversal in the stock sentiment data.

One might ask why the stock sentiment indicator wouldn’t work better by flashing signals upon reaching a band instead of requiring that it reverse below the upper band (in the case of sell signals) or

above the lower band (in the case of buy signals). But doing so could give the indicator a tendency to

be excessively early The sell signal of mid-1989, for example, would have occurred much closer to the bottom than the top had bands been used without the reversal requirement. The effectiveness of using the stock sentiment indicator with the reversal requirement is evident in the real-time results, which include a 65% accuracy rate and a per annum return of nearly five times the buy-hold return.

The Stock Market The sentiment data for stocks produced a

number of good indicators for the stock market, all producing sell signals after reversals from extreme optimism and buy signals after reversals from extreme pessimism. Yet, the stock market’s secular uptrend made it hard for indicators to beat buy-hold, even in the optimization period. In fact, the selected indicator underperformed over that eight-year period. But it did hold up in the test period, beating buy- hold and maintaining a comparable accuracy rate (54%)-as shown on Chart #7, it flashed some particularly timely sell signals in 1987,199O and 1994. If the indicator has a drawback, it’s the high volatility


caused by the short moving average, which can cause it to generate signals upon very short-term moves and thereby gives the indicator a whipsaw vulnerability

The analysis ofgold sentiment as a stock market indicator produced some surprisingly good possibilities. Similar to gold sentiment’s effect on the bond market, its leading inverse influence on stocks can be seen in Chart M-the best indicators flashed buy signals after gold sentiment peaks and sell signals after sentiment troughs. And the relationship can be seen from a longer-term perspective. In 1987, 1990 and in both mid-1993 and early-1994, gold sentiment warned of stock market weakness ahead by rising to extremes. By the time the stock market actually started to fall in each case, the gold sentiment was in a downtrend, followed later by a stock market advance.

The gold sentiment indicator is designed to catch shorter-term cycles within the longer-term moves, and the chart does reveal a tendency to flash signals by catching brief encounters between band and smoothing. This worked to the indicator’s advantage prior to the 1987 crash but to its detriment prior to the 1990 sentiment spike and final run-up to

the market’s 1990 high. As shown in the test period results, the indicator underperformed the market. However, more than two-thirds of its trades were profitable. So on balance, the gold sentiment can be considered useful as an indicator for the stock market.

Of the nine relationships examined, the final one-bond sentiment versus the stock market-was the only one that wasn’t effective to some degree. Several indicators emerged from the optimization, producing signals in the same manner as the stock sentiment indicator. But they simply did not hold up in the test period, rather underperforming buy-hold or failing to post an accuracy rate of more than 50%. The problem with using bond sentiment to call the stock market could be that the cycles are just too out of sync-bond sentiment leads the bond market, and the bond market leads the stock market, which is also led by the stock market sentiment.

The Composite Approach In reviewing the results for the nine relation-

ships it can be seen that each market’s own sentiment can be used to produce effective signals for that market. The intermarket indicators are generally

Wsekly Data l/09/67 - I/06/95 (LO6 Scolr) 3”> 472

440

411

Profitabk Trader: 68% Gain PN Annum: 5.7% Buy-Hold GPA: 6.8% Signal Dates: 1 l30/07 - 116195

I 383 l-

236 -

Bmda = 1.5 Standard Deviatbns From 7-W& Mean Buy Si9nah = Smoothho Moves Above Upper Band & Thar B&w It SalI Sbnab = 9mmthho Movss Below Lower Band I Thm Above It

- 411

- 383

- 357

- 333

- 311

- 290

- 271

- 253

- 236

- 69

48 -

45 - 45

42 - - 42

39 - - l 39 36 - Warnings of Weakness

36

I I

(8) Market Vane Gold Sentiment -- 18Week Exponential Moving Average


effective as well, but to varying degrees, with some behind this composite, since it is simply the combi- more reliable than others. We’ve seen that in gen- nation of the three previously-selected indicators, eral terms, the sentiment for the stock and bond Even so, it generated profits on 55% of its 47 trades markets has an inverse relationship with gold, while while beating the -0.9% buy-hold return with a per gold sentiment has an inverse relationship with the annum return of 8.1%. That’s far better than the stock and bond markets. By way of contrast, the stock per annum returns produced by any of the individual sentiment can be useful aa an indicator for the bond indicators, as summarized in the table. (The com- market, while bond sentiment has no such luck with plete signal record is shown in A.pperr.&x A.) the stock market. In a composite, the majority rules, so a question-

So what is the best way to use these indicators? able signal from one of the three indicators is almost One approach is to identify the most reliable single always ruled out. In mid-1989, for example, the indicator for each of the three markets, and to then indicators based on the sentiment for gold and the use that indicator exclusively Another approach is stock market quickly overruled a poorly-timed sell to combine the indicators into composites for the signal from the bond sentiment indicator, resulting markets they were built to call. As shown in Table in a profitable trade. At times, all three indicators #2, the second approach would have produced the may be on the wrong side of the market, as exempli- best returns in the test period. In the composites fied by the sell signal of early 1993. But as suggested developed for this paper, a buy signal requires that by the composite’s real-time track record, the odds at least two of the three indicators are on buys, are against that happening. whereas a sell signal requires that at least two of the The benefits of using the composite approach are three indicators are on sells.10 also illustrated in Chart #IO, which features the

Chart #9 shows the combination of the indica- composite for the bond market. In late 1989, for tors that use the sentiment for gold, bonds, and example, the stock sentiment indicator generated a stocks to call the gold marketa composite for the buy signal near the bond market top, but the bond gold market. There was no direct optimization and gold sentiment indicators overruled the signal

GoM vs. Composite of Market Vane Sentiment Weekly Data l/09/87 - l/06/95 (Log Scale)

544 - Buy = At Least Two hdhxtom on 5uys - 544 Sell = At Leest Two lndiutom on Sob

- 489

317 - weamtcontmot)

- 80

Msfket vane aonde (15-W& EMAI

- 60

Market Vane Stocks (1 S-Week SMA)

- 60

- 57

- 54

MTA JOURNAL / WINTER 1994 - SPRING 1995 2 1

Bonds vs. Composite of Market Vane Sentiment Wdly Dota l/09/67 - l/06/95 (LOS Scale)

I3M - Profitable Trades: 62%

125 - Gain Per Annum: 9.2% Suy-AtLeastTwoMMorsm9uyr

114 - Buy-Hold GPA: -0.1% SeI=AtLeaatTwolndkatomonSdlr

signal Dates 1123107 - 116195 8

103 -

(13-weak Pafpetual Contract)

15 - - Msdcet Vms Gold

15

122-w.& !wA)

Muka VW Bonds - 65 (14-We& EMA) - 60

- 55

B

Market Vane Stocb _ 10 l&Week WA) - 65

- 60 - 55

25 - B - 25

(10) MM J S N J9fd8bl J 9 N J&M I S N J9;oM J S N J9f M J S N J9F2M J S N ;$3M J S N ;&qM J S N 1 1995 _

by remaining on sells. The bond market subsequently peaked. Like the gold composite, the bond composite outpaced the buy-hold return with a per annum gain of more than 8%, substantially better than the per annum returns of the individual indicators. And it did so with a 62% accuracy rate. (The complete signal record is shown in Appendix B). The gold and bond composite results both illustrate how indicators that are useful on a real-time basis can produce even better real-time results when combined together into a composite.

Conclusion In conclusion, this paper has shown that effec-

tive indicators can be produced using not only a market’s own sentiment, but also the sentiment in other markets as well. It has also illustrated that some relationships lend themselves to more effective inter-market sentiment indicators than do others, and that each relationship should thus be considered for its relative degree of validity.

The foregoing results suggest independence among sentiment indicators, and thus they can be added together for superior results. This independence/additive aspect may be the great advantage of

using intermarket sentiment. It reflects the sentiment of different pools of players, whereas a single market’s sentiment indicators may tap just one pool.

The paper has demonstrated further that during the research process, it is essential to stay close to the basic relationships between the sentiment data and the price data. With excessive data manipulation, those relationships can get lost, and the first real-time signal from the over-optimized indicator might raise the question “so what is this indicator really saying?” If the question cannot be followed by a confident response, such as “we’ve seen a peak in optimism, which is bearish for the market,” then the indicator should be reconsidered.

The analytical methods should thus be scruti- nized. In the case of volatility band indicators in particular, the formulas must be examined closely, and the indicators studied visually, to be confident that they will prove to be reliable in the future. With so many variables in each indicator’s composition, and with so many combinations for the computer to choose from in determining the most hypothetically profitable results, the dangers of overfitting and over- optimization must be taken seriously-i.e., no matter how good an indicator’s hypothetical track record,

22 MTA JOLJRNAL / WINTER 1994 - SPRING 1995

APPBNDIXA

SIGNAL BREAK-DOWN FOR TEST PERIOD: CALLING GOLD WITH THREE-INDICATOR SENTIMENT COMPOSITE (ALL RESULTS INCLUDE TRANSACTION COSTS)

MARKET : GOLD -- Nearest Contract SIGNALS: sentiment Composite FRICTION: 0.250 Percent MARGIN: 100 percent DATES: 2/13/07 through l/06/95 (Weekly651

ACTION DATE

Short 2/13/07 Long 3/20/07 Short 5/01/87 Long 6/19/07 Short 7/3l/07 Long g/04/07 Short l/Oa/00 Long 2/26/0a Short 3/25/00 L-g 3/31/00 Short 4/22/00

Long s/09/00 Short l/13/09 L-g l/20/09 Short 3/03/09 Long s/05/09

short 7/21/09

Low3 g/22/09 short l/05/90 Long 4/27/90

Short a/24/90 Long 11/09/90

3hort 2/01/91

bong 6/07/91

3hort 7/26/91 Long g/20/91 3hort ll/Ol/9l ;ong l/31/92 Ehort 3..;;$;; bong 3hort 7/31/92 iong a/20/92 Short g/25/92 bong 10/02/92 ;hort 10/16/92 ,ong 10/30/92 Short 12/10/92 Long 12/31/92 Short 3/05/93

PRICE ACTION DATE PRICE PROFITI DAYS $10,000

396.01 Cover 407.32 Sell 455.96 Cover 446.11 Sell 462.94 Cwer 464.26 Sell 482.49 Cover 430.67 Sell 452.07 Cover 455.54 Sell 449 -77 Cwer 422.25 Sell 401.29 Cwer 409.52 Sell 305.63 Cwer 370.64 Sell 371.47 Cwer

367.62 Sell 405.20 Cover 372.03 Sell 413.96 Cwer 305.76 Sell 365.10 Cover 367.22 Sell 365.20 Cover 349.37 Sell 354.71 Cover

357.09 Sell 340.63 Cover 339.05 Sell 356.51 Cover 340.75 Sell 340.43 Cover 340.97 Sell 341.54 Cwer 340.25 Sell 336.06 Cwer 333.93 Sell 329.57 Cover

3/20/07 5/01/07 6/19/07 7/31/07

'l;xi;:;: 2/26/88

:;zi 4/22/00

407.32 -2.65 455.96 11.94

siosiaa

:::ig 3/03/89 sio5ies

7/21/09

91;::;98:

4/21/90

e/24/90 n/09/90

2/01/91

6/07/91

7/26/91

g/20/91

11/01/91

l/31/92

~;K;x~

7/31/92

e/20/92

9/2s/92

10/02/92

10/16/92

10/30/92

12jl0j92

12/31/92

3/05/93

10/01/93

446.11 2.16 462.94 3.77 464.26 -0.20 402.49 3.93 430.67 10.74 452.07 4.97 455.54 -0.77 449.77 -1.27 422.25 6.12 401.29 -4.96 409.52 -2.05 305.63 -5.03 370.64 1.01 371.47 -1.09 367.62 1.04 405.28 10.25 372.03 0.21 413.96 11.27 305.76 6.81 365.10 -5.33 367.22 -0.56 365.20 -0.53 349.37 4.36 354.71 1.53 357.09 -0.90 340.63 -2.59 339.05 2.52 356.51 4.90 340.75 4.42 340.43 2.25 340.97 -0.16 341.54 -2.13 340.25 0.30 336.06 -1.23 333.93 0.63 329.57 -1.31 355.09 -7.90

35 9.735 42 10,090 49 11,133 42 11,553 35 11,520

126 11,972 49 13,250 aa 13,917

6 13,010 22 13,635

140 14,470 126 13,751

7 13,469 42 12.684 63 12,914 77 12,669 63 12,000

10s 14,112 112 15,270 119 16,991

77 10,149 a4 17,101

126 17,005 49 16,995 56 17.735 42 10,007 91 17.045 3s 17,303 56 17,021 91 10,694 20 19,521 20 19,960

7 19,929 14 19,505 14 19,579 49 19,330 13 19,460 64 19,206

210 17,673

ACTION DATE PRICE ACTION DATE PRICE PROFIT% DAYS $10,000

Long 10/01/93 Short 11/12/93 Long u/19/93 Short l..,/3;9/;3 Long Short 3/11/94 Long 3/25/94 Short 6/24/94 Long 12/02/94

355.09 Sell 11/12/93 372.17 4.57 42 372.17 Cover l1/19/93 370.94

10,401 -1.02 7

370.94 Sell 12/10/93 302.24 10,145

0.07 21 302.24 Cover 2/25/94

10,302 379.55 0.71 77

379.55 Sell 3/11/94 10,432

304.34 1.26 14 384.34

10,664 Cover 3/25/94 391.98 -1.99 14

391.90 Sell 6/24/94 10,293

390.12 -0.47 91 390.12

18,207 Cwer 12/02/94 376.44 3.51 161

376.44 (Open) l/06/95 10,045

370.27 -1.64 35 10,536

NARKST: GOLD -- Nearest Contract ----lL-

SIGWALS: Sentiment Compoe~ce FRICTION: 0.250 Percent NARGIN: 100 Percent DATBS: 2/13/07 through l/06/95 (WeeklyCS)

Total Number Profit/ Number Profit Trades Trade

Lola2 l--YE

LOSSBS -27.54 11 -2.50 GAINS 61.52 12 5.13 Net 33.97 23 1.48 1353

SHORT LoSSB9 -19.15 10 -1.92 GAINS 53.40 14 3.01 Net 34.25 24 1.43 1496

mSSES -46.70 5 -2.22 GAINS 114.92 4.42 N e t 60.22 47 1.45 2849

Suw4ARY OF CIOSED TRADES

Profitable Tradea: 552 (26 out of 47)

Profit/ Annum

8.76

0.10

0.46

tGein/~Gafn+2Loee ( 71.1&j $Gein/$Gein+$Losa ( 68.3t) Maximum Drawdown Was: -23.10 (7/30/93)

SGain/Loss

RESULTS OF ALL TRADBS (Closed + Open)

$10,000 became $10,536 in 2804 days ( 7.90 years). 0.1% per annum compounded annually.

( 2.2))

BUY/HOLD is -0.9t per annum compounded annually for 2004 days ( 7.90 years).

APPENDIX B

SIGNAL BREAX-DOWN FOR TEST PERIOD: CALLING BONDS WITH THREE-INDICATOR SENTIMENT COMPOSITE (ALL RESULTS INCLUDE TRANSACTION COSTS)

MARKET: BONDS -- Perpetual ContraCt MARKET: BONDS -- Perpetual Contract SIGNALS: Sentiment COmpOeite SIGNALS: Sentiment Composite FRICTION: 0.250 Percent FRICTION: 0.250 Percent MARGIN: 100 Percent MARGIN: 100 Percent

DATES: l/23/07 through l/06/95 (Weekly651 DATBS: l/23/07 through l/06/95 (Weekly65)

ACTION DATE

Short l/23/07 Long S/01/07 Short 7/17/07 Long g/25/07 Short 3/U/00 Long S/13/00 Short W2,D:;; Long Short ;$2;g;0 Long Short :$:b;;; Long Short 6/09/09 Long 3/02/90 Short 4/20/90 Long Silli Short C/29/90 Long 10/05/90 Short 3/15/91 Long 7/19/91 Short 10/10/91 Long 11/00/91 Short 2/21/92 Long SiOli92 Short S/22/92 Long 6/05/92 Short 6/12/92 Long 7/10/92 Short g/04/92 Long 10/23/92 Short ;{2,;:;: Long Short 10/22/93 Long 12/03/93 Short l/07/94 Long l/14/94 Short 2/11/94 Long 4/00/94 Short 7/00/94 Long 11/11/94

PRICE ACTION DATE PRICE PROFIT% DAYS $10,000

99.66' 91.40 91.65 82.39 92.05 07.15 04.51 07.66 90.21 90.54 08.59 09.19 96.54 93.23 09.10 92.29 94.05 91.04 94.33 93.45 90.53 99 -94 90.44 90.31 99.72

100.00 99.34

102.22

Cover Sell Cover Sell Cover Sell Cover Sell Cover Sell

Sell Cover Sell

Sell Cover Sell Cover Sell Cover Sell Cover Sell Cover Sell Cover Sell

5/01/07 7/17/07 g/25/07 3/11/08 s/13/00 5/20/00 6/10/08 10/20/00 l/13/89 2/10/89 4/14/89 6/09/89 3/02/90 r/20/90 5/11/90 6/29/90 lo/OS/90 3/s/91 l/19/91 lo/la/91 ll/O8/91 2/21/92 5/O1/92 5/22/92 6/05/92 6/12/92 7/10/92

91.40 0.21 91.65 0.10 02.39 10.10 92.05 11.72 07.15 5.32 04.51 -3.04 07.66 -3.73 90.21 2.92 90.54 -0.36 00.59 -2.15 09.19 -0.60 96.54 0.24 93.23 3.43 09.10 -4.34 92.29 -3.49 94.05 1.90 91.04 3.20 94.33 3.61 93.45 0.93 98.53 5.44 99.94 -1.42 98.44 -1.50 90.31 0.14 99.72 1.44

100.00 -0.20 99.34 -0.65

9jo4j92 10/23/92 3/26/93

102.22 -2.90 105.73 3.44 102.10 3.44 100.63 6.40 114.30 -5.29 110.00 3.23 115.29 2.36 115.46 0.15 115.29 0.15 114.06 -1.07 104.79 0.12 100.06 -4.51

95.06 4.20 99.03 3.31

90 77 70

160 63

21 140

77 20 63 56

266 49 21 49 90

161 126

91 21

105 70 21 14

7 20 56 49

154 105 105

42 35

7 20 56 91

126 56

10,021 10,040 11,935 13,334 14,043 13,616 13,109 13,491 13,442 13,153 13,064 14,140 14,624 13,989 13,501 13,750 14,190 14,710 14,046 15,654 15,431 15,200 15,221 15,439 15,396 15,295 14,052 15,362 15,090 16,907 16,013 16,531 16,922 16.947 16,973 16,792 10,156 17,336 10,064 10,661

105.73 102.10 100.63 Cover ljO9j93 114.30 Sell 10/22/93 110.00 Cover 12/03/93 115.29 Sell l/07/94 115.46 Cover l/14/94 115.29 Sell 2/11/94 114.06 Cover 4/08/94 104.79 Sell l/08/94 100.06 Cover 11/11/94

95.06 (Open) l/06/95

Cover Sell

Total Number Profit/ Number Profit Trades Trade Days

LONG LOSSBS -17.27 7 -2.47

GAINS 40.66 12 4.06 Net 31.40 19 1.65 1420

SHORT LOSSES -10.14 0 -2.27

GAINS 49.50 12 4.13 Net 31.43 20 1.57 1421

TOTALS MSSES -35.41 15 -2.36

GAINS 90.24 24 4.09 Net 62.03 39 1.61 2049

Profit/ Annum

7.05

7.09

7.07

SUbMARY OF CLOSED TRADES

Profitable Trades: 622 (24 out of 39)

*Gain/%Gain+ZLoes 1 73.5%) $Gain/$Gain+$Loee ( 71.32) $Gain/Loee ( 2.5)) Maximum Drawdown Wae: -5.51 (10/16/07)

RESULTS OF ALL TRADES (Cloeed + Open)

$10,000 became $10,661 in 2905 days ( 7.96 yeare). 8.2% per annum compounded annually.

BDY/HOLD is -0.1% per annum compounded annually for 2905 days ( 7.96 years)

the results are of little use if the indicator doesn’t work well when actually used for making real-time trading decisions,

The research addressed this issue by optimizing over an eight-year span and then testing over the subsequent eight-year period. Although the approach doesn’t guarantee that any indicator will be fail-safe in the years ahead, it does reduce the risk that an indicator will break down, and it is preferable to optimizing over the full 16-year period without any real-time testing. Another approach would be to re- optimize on a yearly basis, always using the latest four years of data for the optimization and always requiring that the indicator run real-time for a full year before the next optimization.

In the analyst’s tool chest, intermarket sentiment indicators can claim a space alongside conventional sentiment indicators. As more and more markets become increasingly liquid and popular, and as they become increasingly accessible, it should become increasingly necessary to assess the sentiment in other markets when assessing the potential for inflows into, and outflows from, one’s primary market of interest. Intermarket sentiment should also become increasingly important for those who must make allocation decisions and for those who trade various markets. Among the technical analysis questions that warrant continued research, intermarket sentiment deserves high consideration.

caters into composites. One simple approach is to

This paper has also shown that one of the ways

combine two intermarket sentiment indicators with

to use the sentiment indicators to reduce the risk of

an indicator based on the market’s own sentiment.

losing from a bad real-time signal is to combine indi-

The two composites developed for this paper generated returns that were substantially better than the returns generated by any of the individual indicators. This fact cannot be emphasized enough. It argues that intermarket sentiment provides added value, expanding the sentiment perspective, enhanc- ing the effort to identify market reversals, and thus improving the decision-making process.

The next step would be to test the sentiment for numerous others markets-such as the U.S. dollar, CRB Index or T-Bills-and to then expand the number of indicators in a composite or choose the top few from the expanded pool of possibilities. One might also opt to keep a market’s own sentiment as a high percentage of a composite’s contents. Additionally, testing could be done on indicators based on the ratio of sentiment in one market to the sentiment in another. There is a multitude of possibilities for further research in the area of intermarket sentiment.

But whatever one decides. to use-whether they are individual indicators or composites-it is important to remember that even the indicator with the most reliable real-time record can at best reduce risks and help you outperform the markets. It cannot be expected to produce phenomenal profits by itself. And any indicator is of little value if it isn’t used properly, or isn’t used at all. In the same way that a carpenter cannot hope to build a house by using the wrong end of a hammer, or not using one at all, a trader cannot hope to build profits by acting opposite his signals, or simply ignoring them altogether. And trying to time the market without using any indicators would be like trying to build a house with your bare hands. Profit-building can be a lot less difficult with the assistance of proven indicators used properly

REFERENCES

1. Murphy, John J., Zntermarket Technical Analysis, John Wiley & Sons, Inc., 1991.

2. Davis, Ned, Being Right or Making Money, Ned Davis Research, Inc., 1991.

3. Investor’s Intelligence, 30 Church Street, New Rochelle, NY 10801.

4. Consensus Inc., 1735 McGee Street, Kansas City, MO 64108.

5. Market Vane, PO. Box 90490, Pasadena, CA91109.

6. John A. Bollinger, Bollinger Capital Management, PO. Box 3358, Manhattan Beach, CA 90266.

7. It could be argued that since the Market Vane data is reported on a scale of 0% to lOO%, there could be periods when the bands would be above 100% or below 0%, making it impossible for the indicator to generate a signal. But the weekly numbers have never reached those extremes since 1978, instead ranging from 13% to 92% for gold, from 15% to 92% for the stock market, and from 13% to 90% for the bond market. The use of moving averages further reduces the chances that the data would ever move close enough to 0% or 100% to send a band beyond it. However, the possibility should still be recognized.

8. All of the results discussed in this paper are based on using price data only. They do not include dividends for the stock market or interest for the bond market. The results are based on taking long positions on buy signals and closing out those positions on sell signals, at which time short positions are taken and held until closed out by a new buy signal.

9. While this was the most dramatic case of improvement, the maximum drawdown in the test period also improved from the maximum drawdown in the optimization period for five of the seven other individual indicators that emerged from the research.

10. An exception is a composite’s first signal, which is determined by the first of the three indicators to flash a signal.

ACKNOWLEDGEMENT

The author would like to acknowledge Ned Davis Research for the excellent analytical and graphics capabilities of the firm’s computer program, which was vital to the research and charting done for this paper.

Tim Hayes has been Editor of the institutional Stock Market Strategy since 1986, developing indicators, models and studies for the equity and international services of Ned Davis Research, Inc. of Venice, FL. Tim is a member of the Market Technicians Association and former associate editor of the MTA Newsletter. His research articles have been featured in Technical Analy- sis of Stocks and Commodities and other publications.

MTAJOURNAL /WINTER 1994 - SPRING 1995 25

Location, Location, Location by Tamalyn V. Crutchfield

Candlestick patterns are effective tools in identify- Candlesticks are best used for early warning in- ing market reversal points. Candlestick formations dicators as opposed to independent trading tools. typically provide the first indication of a directional Although candlestick reversal formations alert to change, frequently well in advance of traditional directional changes, their effectiveness is limited by indicators. their inability to project price targets. Confirmation

Candlestick charting techniques, used by the from traditional indicators is required before a trad- Japanese for over a century, were formally introduced ing decision is executed. to the West by Steve Nison, CMT, approximately four Traditional technical indicators compare current years ago. The popularity of candlestick charting has market conditions with past conditions to identify evolved in phases similar to current fashion trends. market direction and momentum. These indicators Candlestick methodology was first a chic topic at fall into two categories: trend-following and over- technical analysis forums. Today it is a highly re- bought/oversold. spected form in which to view data. Terms such as Trend-following indicators, such as Moving Av- ‘hanging man’, ‘dark cloud cover’, and ‘o?oji star’ erages, Moving Average Convergence/Divergence are now part of the market technician’s vocabulary. (MACD) and Directional Movement Index (DMI), are Many computer software packages eontain standard- well suited for trending markets. These indicators ized candlestick charting applications. do not function well in consolidating or range-bound

The basic premise of the candlestick chart is markets. based on the open, high, low and close of the market Conversely, overbought/oversold indicators are in a particular time frame, i.e., intraday, daily, weekly, well suited for consolidating or range-bound markets. etc. The range between the open and the close is Their reliability decreases when markets demon- called the real body of the candle. If the close is higher strate strong trends. Examples of overbought/over- than the open, the real body is white. If the market sold indicators are Stochastics, Momentum and close is lower than the open, the real body is black. Relative Strength Index (RSI). The high and the low of the market are attached to Combining any or all of these indicators into a the real body as lines, such as found in a standard trading system, provides a clearer identification of bar graph. These extensions beyond the real body current market conditions. are called shadows. The upper shadow, protracting Our focus is on the following traditional indica- to the session high, is placed above the real body and tors and their relationship to several candlestick the lower shadow, protracting to the session low, is patterns. placed below the real body

Trend-following Indicators: THE CANDLESTICK Moving Average-a lagging indicator that sig-

nals the beginning or end of a trend. A simple moving average is an average of prices over a given

TT period of time.

01

When two or three moving averages are used a

CLOSE OPEN buy signal occurs when the shorter moving average crosses above the longer moving average(s). Con- versely, a sell signal is generated when the shorter

OPEN CLOSE line crosses below the longer line(s). When employ-

f f ing the three moving average system the crossing of

SHADOW the middle line with the longer line confirms the buy or sell signal. The most common three day moving average system uses the 4,9 and 18 day averages.

Moving Average ConvergenceDivergence (MACD) - is made up of two exponentially smoothed moving average lines: the MACD line and the signal line.

A buy signal is generated when the faster (shorter) line crosses above the slower (longer) line. A sell signal is generated when the faster (shorter) line crosses below the slower (longer) line.

Overbought/Oversold Indicators: Stochustics - consists of two lines: %K and %D.

These lines oscillate between 0 and 100. Values above 80 are in overbought territory and values under 20 are oversold.

Signals occur when the %D line diverges with the underlying price and the %D line is in the overbought or oversold zone. Buy when the %K line crosses the %D line after the D line reverses direction from down to up. Sell when the %K line crosses the %D line after the D line reverses direction from up to down.

Relative Strength Index (RSI) - is used tc smooth price movement. RSI is plotted on a scale oj 0 to 100. Values above 80 are deemed overboughl and values under 20 are oversold. Buy and sell sig nals are generated from divergence between the RSI and the underlying price.

Real estate property analysis provides the besl analogy for considering the use of candlesticks. The foremost canon in purchasing real estate is location This aspect is equally true when trading candlestick reversal patterns. The signal may be valid, but it i; important to be cognizant of the point where thf reversal signal appears within the framework of tht chart. The candlestick provides a preview of what iz to come, but it does not provide a definitive target OI timeframe.

The basic parallelism between real estate am candlesticks lies within the candlestick pattern’s geo graphic location on the chart. Visualize a candlesticl formation the same way a prospective house purchase

. . .

. . . , , .

. . * . . .

BUY SIGNAL

FIGURE1 Silver

June 1994


would be considered. The house may have a beauti- ful interior, but the exterior may require extra attention. Additional consideration is given to the surrounding neighborhood in which the structure is located. Purchasing the best house in a deteriorat- ing environment would be a questionable decision. A similar approach is warranted when evaluating candlestick signals. A trader does not want to be placed in a position of perceiving a ‘buy’ signal while all his/her colleagues continue to see a weakening market.

First Alert June Silver, 199” The silver market appears to

be range-bound between $4.90 and $5.60. The daily price patterns consist of steep rallies followed by steep declines. An early February move tests the low and establishes the high of the range. The subsequent reversal retraces approximately 50% of the uptrend and consolidates into a $0.20 trading range.

A second attempt to rally the market upward occurs in mid-March.

March 14-16th provides an example of a candlestick indicator. Figure 1 illustrates the completed evening star candlestick formation.

The evening star pattern requires three periods of data. It begins with a long white candle followed by a small candle formation that gaps above the first candle. The third period consists of a long black candle that closes within the body of the first candle. A 50% or more penetration is pre- ferred. This is considered one of the most reliable reversal indicators.

A closer examination of the terrain is warranted. On March 14th, silver trades two cents shy of the contract high of $5.54. On the close of March 16th the g-day stochastics are approaching overbought valuations at %K=74.24 and %D=71.85. The 5-day RSI value is turned downward at 53.60. The MACD generated a buy signal five days prior

. . .

FIGURE 2

Silver June 1994


FIGURE 3

Silver June 1994

and remains in a buy mode. Examining the land- scape we see a price peak was formed in early February at $5.54. All suggest a top is in the making but none confirm an imminent trend reversal. Looking ahead to March Zlst, a doji forms. A doji occurs when the opening and closing prices are the same. Dojis indicate indecision about market conditions.

Reviewing the supplementary indicators for March 21st in Figure 2: RSI=59.95, stochastics- %K=73.15, %D=71.75 and the MACD values project higher prices. The 4-day moving average has turned downward but the 9- and U-day moving averages remain in an uptrend. The evening star/doji formations may be signals warranting additional scrutiny, but there is no confirmation of a trend change by the traditional indicators.

Looking into the future, Figure 3 indicates that silver continued to rally to $5.82 before

sharply correcting on April 4th. The evening star and doji formations warned of

an impending reversal, but confirmation was not close at hand.

You Make the Call March Treasury Bond, 1993-The treasury

bond market concludes 1992 in a strong uptrend predominated by long white candles. Resistance is established and tested at 105-20 and is then immediately followed by a series of black candles. This corrective action finds support near the 50% retracement point. The ensuing move is quick and bullish, penetrating resistance and driving the price up to 107-16.

Here is an example for candlestick disciples. This rally, portrayed in Figure 4, is interrupted by the formation of a dark cloud couer during Janu- ary 25-26th.


FIGURE 4

T-Bond March 1993

A dark cloud cover is a bearish reversal signal that requires two periods to complete. The first period is formed by a long white candle and the second period consists of a black candle. This black candle opens above the white candle’s high (upper shadow) and closes at or below the midpoint of the white candle.

TABLE I

T-Bond - March 1993

DATE OPEN HIGH LOW CLOSE

l/25 106-04 107-09 106-04 107-07

l/26 107-13 107-20 106-20 106-21

The midpoint of the white candlestick is 106-21+ and the close of January 26th is 106-21-exact midpoint penetration. Do you sell the bond at this level or wait for additional confirmation?

Additional assistance is needed to examine the surrounding terrain. The g-day stochastics are in

overbought territory at %K=85.20 and %D=85.39. The %K line has turn down to the %D line. How- ever, the %D line has not changed directions, thus, these values do not confirm the candlestick reversal signal. The RSI is 67.52 and the 4,9, and 18 day moving averages are 10616,105-28 and 105-07, respectively These signal an uptrending market. We should not postulate that the January 26th high of 107-20 is the termination point of this move.

Figure 5 reveals what actually transpired. The reversal signal was rejected, and the market continued higher. The traditional indicators did not provide confirmation of a reversal.

The Perfect Combination December Gold, 1993-Although it is recommended

that supplemental indicators be consulted before a trade execution, it is not implied that candlesticks are not viable trading tools. In this analysis the gold


FIGURE 5

T-Bond - March 1993

Dai) [14122193 W10 +IItI O=lWlt E=lWl2 MIY’O5 bhdvg3hes MI2 lK16 lI@l

.

DJiRK C,LOUD,COV&R\

.


r market is in a steady bullish trend. The unvary- ing corrective move fills a gap from the prior month and the bull rally continues its measured ascent.

Figure 6 illustrates a frequent occurrence where the candlestick patterns and traditional indi- caters align. On July 6 and 7, a ‘doji starhanging man’ combination is formed.

A doji star is a candle with the same opening and closing prices that gaps above the preceding white candle or below the preceding black candle. The hanging man is a candle with a small real body, little or no upper shadow, and a lower shadow mea- suring at least twice the length of the body. These formations independently are very bearish.

Upon further perusal of the market, we observe new highs and a topping RSI value of 89.19. The stochastics are near the upper boundary of the overbought range at %K=94.59 and %D=89.65. The Elliott Wave count is incomplete, but implies that this is a third (III) wave. In this instance a fourth

(IV) wave projects lower prices. A reversal appears imminent.

Expanding the chart to view the market’s response shows a perfectly correlated reversal signal. In Figure 7, the traditional indicators confirm the candlestick reversal signal the next session. A stochastics sell signal was generated at the point the %D line turned downward and was crossed by the %K line. The RSI reversed direction and the fourth (IV) wave of the Elliott Wave sequence commenced.

Convergence of the various indicators brings cre- dence to their reliability.

Conclusion Candlestick patterns are effective tools for ana-

lyzing market action. They are particularly helpful in identifying market reversal points and trend changes. Candlesticks have limited use as technical forecasters because they do not project price targets, merely trend changes. Trading exclusively using

/


FIGURE 7

Gold December1993

candlesticks requires waiting for reversal patterns on the entry and exit sides without definitive objectives. A cardinal rule of futures trading is ‘timing is everything’. Since the entry point can make or break a trade, this is an ineffxcient (and possibly unprofitable) method of trading. It is pertinent to employ several traditional Western techniques as ‘timing indicators’.

BIBLIOGRAPHY

Murphy, J.J., Technical Analysis of the Futures Markets. New York, NY: New York Institute of Finance, 1986.

Nison, S., Japanese Candlestick Charting Techniques. New York, NY: Simon & Schuster, 1991.

Wagner, G.S. and Matheny, B., ‘Candlesticks and Intraday Market Analysis’, Technical Analysis of Stocks and Commodities (April, 1993).

Wagner, G.S. and Matheny, B., Trading Applications OfJapanese Candlestick Charting. New York, NY John Wiley & Sons, Inc., 1994.

Tamalyn V Crutchfield is the President and founder of TVC Trading Co. The firm trades commodity interests on listed futures and futures options contracts. Ms. Crutchfield is a frequent lecturer at the New York Znstitute of Finance. Ms. Crutchfield received an MBA from Wharton, University of Pennsylvania, a B.S. in Chemical Engineering from Georgia Znsti- tute of Technology, and a B.S. in Chemistry from Spelman College.

AI1 the charts in this paper were created by the TradeStation software program, Omega Research, Miami, FL.

The data for the charts was supplied by Signal, Data Broadcasting Corp., San Mateo, CA.


r

System Testing for Consistent Profitability by Muneer Al Hulaibi

Introduction The Ingredients Optimization of trading rules is generally under- To try out my CAP testing method I decided to stood to mean testing the rule over the largest use what is probably the simplest trading rule ever: amount of historical data available, all in one block. buy when the low of the day crosses over the moving The resultant optimal variables generated would average, sell when the high moves below the moving then be considered for the next year of trading. I average. This rule would keep me in the market all believe there is a more effective method of testing the time. The only variable to test over the data was trading rules. the moving average period.

We usually report trading results at the end of It should be noted here that it is really immate- the year. If we have used a particular trading sys- rial which trading rule we use. What we are trying to tern during the past year which at least met our investigate here is the CAP testing method for test- profit target, then we would probably be inclined ing trading rules. We are not investigating the trad- to use the same trading system for the next year. ing rule itself

However, if our trading system fell short of the It may be argued that the highs and lows of target or made a loss we would be more inclined to the day are only known after the close. The ques- abandon it for the next year’s trading. We would tion arises, “How can we deal at the close after find it difficult to justify its use to ourselves or to the close?” This really is subject to what our deli- our employers. nition of the close is. If what we mean by the

With this in mind, I find it difficult to understand “close” is the literal close (i.e., the last price of the why traders optimize a rule over a period of say forty day), then yes; we would probably not be able to years and use the variable which came up with the execute a deal at that price but would have to wait maximum profit for the next year’s trading. This till the next day’s open. same optimized rule may have shown a loss in a num- In many cases price data that is collected uses the ber of years out of those forty. Picture yourself some- prices that is quoted at a certain time of the day as where in the middle of those historical forty years in being the “close”. This is certainly the case in the twenty an area where this optimized rule made losses in three four hour foreign exchange markets where there is no consecutive years. Picture yourself explaining to the real close. Bar charts plotted for those markets usually boss why you should stick with this rule because after indicate the close as being the price quoted at 5.00 p.m. completing forty years of trading it will have made eastern time in the U.S.A. Trading does not stop at the maximum profit. that time. A European trader might choose 5.00 p.m.

Since we are expected to report annual profits, U.K time as his or her closing price. then it would make sense to test a rule for annual For analysis purposes we need to plot and use maximum profits. Testing the rule for a block of forty a price to factor in our trading model. The selec- years of historical data would make sense only if we tion of this “close” will become more and more a were going to be held accountable for our profits subject of opinion and personal circumstances as another forty years down the road. stock and commodity markets become more global.

My aim here is to explain the Consistent Annual Therefore for now and for the purposes of this in- Profitability (CAP) testing method. CAP will test a vestigation of the CAP testing method we shall con- trading rule, any trading rule, for blocks of one year tent ourselves that it is possible to place a buy or each. The CAP testing method will reveal the vari- sell order at the close. able or variables which perform consistently well every I used Computrac software to test the trading single year out of a total batch of many years. Out of system. these consistently profitable variables we will use the I decided to use the commodity futures data one which achieved the maximum total profit in all provided by Tick Data. The commodity I chose was the years for the next year’s trading. the British Pound. There was no specific reason for


choosing this commodity I could have easily chosen any other.

Using Tick Data’s data management software, I created one long continuous file of British Pound prices stretching from 1975 to 1993. This was done by joining several contracts together. When one contract expired the one with the next expiry date was “spliced” in.

When creating such a contract there is of course the problem of a price jump in the data where the rollover takes place. To overcome this problem and give a continuous smooth price move the price spread on the rollover date between the expiring contract and the rollover contract was artificially removed. The price difference was added (or subtracted) to every subsequent price in the file. The effect was cumulative over subsequent rolls.

In this investigation I have deliberately chosen to exclude all frictional costs, such as commissions or slippage. This I did for the sake of sim- plicity Spreads, broker fees, commissions, slippage, dividends, interest costs and any other expenses can vary widely Spreads can vary depending on the amount traded, the relationship between the trading parties, the volatility of the market at the time of the trade, the “position” that the seller is holding, i.e., is he or she already long or short of the commodity. Broker commissions vary too. There are many discount brokers around now and there is a lot of competition in fees charged. Brokers may charge different fees for different amounts traded. There may by volume discounts. Some traders, such as banks for instance, may deal direct with a counterparty and avoid a broker entirely

Different markets will have different trading costs. The CAP testing method may be applied to different markets. I have used futures data in my illustration but am by no means restricting the application of CAP to futures trading.

The point I am trying to make here is that trading costs are specific to each trader’s circumstances. In using CAP a trader will have to plug in his or her own trading cost allowance.

Recipe I loaded the data in portions of two years into

the Snap module of Computrac. I tested each year from 1976 to 1987 separately I optimized the moving average period for each year using period values of 1 to 100. Positions that were still open at the end of the year were closed at the closing rate for that year. In real life this is what would happen. An open position is squared in the books at the year’s closing price. Then the position would be carried over to the new year at the end-of-year closing price.

Although I loaded two years of data at a time I

only tested one year This extra data was necessary to allow for the lag periods of the moving averages. The system only started trading in the new year being tested. Testing for single years is the basis of the CAP method.

There was no specific reason for using 12 years for testing. I just thought a dozen years sounded reasonable enough. After the results of these twelve years have come I would have my conclusions applied to the rest of the period, namely 1988 to 1992.

First Impressions The next step in the CAP testing method is to

analyze the results of the optimization runs. After the optimization runs were complete the

following table of results was drawn. Please note the results are in terms of U.S. dollars profit per British Pound traded. For example, in 1976 the average profit per British Pound traded, using the 10 day moving average period, was $0.1305.

In Figures 1A and 1B the highlighted figures show the highest profit for that year that any moving average period could achieve.

The top profits for each year (Figure 2) were as follows:

I I 1

Figure 2 Highest Gaining

Moving Average Periods (MAP)

1980 14 0.4!m

1981 10 0.7545

1982 2 0.2m

1983 8 0.2125

I I TOTAL 4.2935

I ’ The profits for every year 1976 through to 1987

using the ideal moving average periods were added to yield $4.2935 per British pound traded. A nice profit if1 had known before the beginning of each year which moving average period to use. However, real life is rarely this kind to one.


Figure 1A Moving Average Period (MAP) Profits (l-50)


Figure 1B Moving Average Period (MAP) Profits (50-100)

AdA@] 1976 1 19W 1 1978 1 197!0 1 19W 1 I981 1 1982 1 1983 1 IW 1 1BS 1 I986 1 1987 5f 0.1475 1.. eJg&] 0.3034 I 0.14a I 0.165o I 0.3610 I -o.am I 0.0320 1 0.1750 I 0.01oo 1 0.01m I 0.3195

52 0. I475 [::,.&g.$;r -

0.m 1 O.Iyo 1 0.1850 _ 1 0.3610 1 -o.ooos 1 0.0510 1 0.175o I O.Oloo ] O.OlP 1 0.3195

0.2140 1 o.moI 0.1850 1 0.3810) o.am 1 0.0510 1 0.17% 1 0.0100 1 0.0120 1 0.3195

L L 0. - 165O

- q.0590 0.4Mo

d . . . . . . 0.1920 1 4m35oI 0.1490 1 0.4510 1 0.0955 1 0.1960 1 o.lm 1 0.1~~ 1 41m BFj

._... 0230 1 1 *

-0.0350 0.1490 I 0.4510 i.....

0.0955 0.1160 0.1660 O.]2p -0.m .$$

: O.luo I 0.09-m I 0.11

0.1610 1 O.ouO 1 0.1190 [ 0.5060 1 0.1745 1 0.1550 1 0.1360 1 0.1100 1 -0.03m 1 0.3605

0.1610 1 0.06401 0.1190 1 0.1745 1 0.1550 1 0.13tW I 0.05lOI -O.OXIO! 0.3495

0.1610 I 0.0590 I 0.1170 I 0.5060 I 0.1745 I 0.15% I O.lm I 0.0510 I 0.030 I 0.3495 I

0.1610 1 0.0030I O.lOdOI 0.5060 1 0.1745 I 0.15501 0.1360~

0.1010 o.al9o O.oQID

o.om o.oao 0.m

O.ooO 0.0090 O.oQID

a

o.ooo -0.0580 O.oQIo

0.08oo -0.05m 0.0970

o.om 0.06m 0.09m

0.08m 0.05m 0.0730

0.5o55 I 0.1nr I 0.0980

0.0980 3 0.0990

O.lpo

0.1490

=I 0.1650

0.1960

0.06m

=I 0.0610

0.0610

0.01 lo =I

0.m

0.m

0.3025

=I O.Wl.5

0.3025

0.5055 O.lU5

* 0.5o55 0.1645

oso55 0.1645 o.oom I O.Wk5

4 0.3025

0.3025

0.1200

-t O.lo#)

0.0350

-t 0.0380

0.0030 -I 0.W

0.4325 0.1515

ti

0.4325 0.1515

0.4325 0.1515

0.0380

=I O.lDO

o.lm


You will notice from the results in Figure 1A year was good for all moving average periods. Even that I included moving average period 1. Moving the worst performing one, moving average period average period 1 is not really a moving average. 35 scored $0.1295. A look at the price chart for Using a period of one means that the price itself is 1981 (see Figure 3) showed me why this was so. used in the calculation (the average of one price is The prices trended in a very nice straight line from the price itself). However, I thought the results January to late September from the 2.7300 level would be interesting to note. In this case we bought down to the 1.9200 level. A trend following sys- when the low moved above the previous day’s close tern such as this one tends to perform well in this and sold when the high moved below the previous kind of market. days close. As you can see from Figure 1A this mov- The lowest annual top profit was recorded in 1986 ing average period was not reliable at all and produced with moving average period 2 yielding $0.1825 per a loss in most of the years. British pound traded. 1986 was a bad year. 71 out of

I could see immediately that moving average the 100 moving average periods tested made a loss in period 2 was the highest gainer in years 1982,1984 that year. The worst drawdowns were also recorded and 1986. That was three years out of the twelve. All in that year by moving average periods 35,36 and 37 the other highest gaining moving average periods only at -80.1630. shined in one year each. This almost tempted me to A look at the price chart for 1986 (see Figure think that I had found my ideal moving average pe- 4) revealed the reason for this bad performance. riod. After more examination I found that moving Prices in 1986 zigzagged up and down for practically average period 2 made losses in 1978,198O and 1987. the whole of that year. Trend following systems This means that the moving average crossover tend to perform extremely badly in consolidating method as it is traded here using period 2 for British markets. pound futures has failed the CAP test. It was interesting to note that the highest gainer,

The highest single profit was recorded in 1981 moving average period 10 for 1981, which was the with moving average period 10 yielding $0.7545 per highest single year profit in all of the twelve years British pound. As a matter of fact that particular was not a consistently reliable moving average

Figure 3 British Pound 1981

2.8

2.7

2.6

2.4

a

2.3

2.2

2.1

2

1.9

PeWI Aprbl


Figure 4 British Pound 1986 1.7

is

1s

is4

162

lb

k 159

136

154

152

15

1.48

1.46

period. It made a loss in 1986. Therefore it failed the CAP test.

I also noted that it was dangerous to assume that a moving average period that performed well in one year would perform well in the following year. Some testing methods make this claim. How- ever, from the chart I could see that moving average period 2 which was the highest gainer for 1986 made a loss in 1987.

Moving average period 2’s poor performance in several of the years also proved another point: a moving average period which performed best in the most difficult markets (in this case 1986) would not necessarily perform well all the time.

The next step I took was to compute the totals of the performance of each moving average period for the years 1976 to 1987.

Revelations The first thing I noted was that all moving aver-

age periods made a total profit for the years 1976 to 1987 (see figure 5). This proved that this trading rule, the moving average crossover method was a reliable rule in terms of profitability no matter which averaging period was used. (Moving average period 1 does not count since it is not a moving average).

Figure 5 Moving Average Periods’ Performance 1976-1987


Interestingly, moving average period 13 was the highest total gainer although it was not the highest gainer in any individual year.

The 7 top gainers formed a cluster on the dia- gram. They were moving average periods 11 through 17. The fact that the top gainers were huddled together confirmed the stability and reliability of the trading rule. A more erratic distribution of profitability with the values of the moving average periods would have made this trading rule suspect and difficult to bet my money on.

Figure 6 Surviving Moving Average Periods (MAP)

of the 1976-1987 Test

Moving average period 13 did not make a loss in any individual year. Its highest profit was in 1981 (the best year for all moving average periods) with $0.5825 profit. Its worst year was 1984 with $0.1030 profit per British pound traded. We can conclude that period 13 passed the CAP test.

This brought me to my next important stage. I decided to discard any moving average period which made a loss in any year. The criterion was to choose the rule or rules which consistently performed the best. That left me with the following moving average periods and their total profits (see Figure 6).

My conclusion at this stage was that moving average period 13 was the best to use. It never made a loss in any year and was the highest gainer of the survivors. It was the favoured variable in this CAP test.

The next question on my mind was, ‘Will it perform equally well in the next few years?”

The Moment of Truth Now came the real challenge for CAP Would the

conclusions of the CAP test prove themselves in the next period of trading?

To see what would have happened, I tested for profitability using the moving average periods that survived for the years 1988 through to 1992. I produced the following results in Figure 7.

Moving average period 13 was still the top performer. Moving average period 15 had fallen out as it made a loss in 1989. Moving average periods 85 to 88 also dropped out as they made losses in 1988 and 1989. The survivors were now moving average periods 9,11,12,13,14 and 21.

We can therefore see that the CAP test here proved its findings to be reliable. Period 13 was consistently profitable for the next live years of trading.

Figure 7 Surviving Moving Average Periods’ (MAP) Performance 1979-1992

do MTA JOURNAL /WINTER 1994 - SPRING 1995

Figure 8A 1986-1990 Moving Average Period (MAP) Profits (l-50)


Figure 88 Moving Average Period (MAP) Profits (51-100)

t- 1986 1987 1988 1989 1990 TOTAL

51 0.0180 0.3195 0.1183 -0.0784 0.2172 0.5946

52 0.0180 0.3195 0.1211 4.0784 0.2172 0.5974

53 -0.0120 0.3195 0.1003 -0). 1040 0.2136 0.5174

54 -0.o2oo 0.3145 0.1003 -0.1200 0.2424 0.5172

57 -0.0 F3= 58 -0.0 59 -0.0

I I

661 -0.1

180 0.3 180 0.3

Ml 0.3

t40 0.3

MO 0.3

I .lO 0.3 .a 0.4 .60 0.4 .70 0.4 i70 0.4 I 681 671 -0.1220~ -0.094ol 0.4175 0.4175 [ I 0.1113 0.0761 1 I 0.2076 -0.2380 1 I 1 0.2576 0.2576 0.4264 0.44%

I I

69 -0.1140 0.3805 0.0753 il.2612 0.2576 0.3382

70 -0.1060 0.3805 0.0471 -0.2624 0.2576 0.3168

71 4.1060 0.3805 0.0301 -0.2624 0.2064 0.2486

72 -0.1060 0.3805 0.0301 -0.2624 0.2108 0.2530

79 -0.0320 0.3495 0.0161 -0.11% 0.1912 0.4052

80 -0.0320 0.3495 0.0161 -0.11% 0.2164 0.4304

811 -m3oo 0.3525 0.0161 4.1124 0.2164 0.4426 t 82 83 1 I -0.0570 -0.0630 0.3485 0.3485 1 I -0.0129 -0.0129 -0.0&a -0.1260 0.2380 0.2696 0.4562 0.3906

\ I I I L

84 -0.0110 0.3485 -0-0189 -0.08m 0.2696 0.5a22

85 0.0110 0.3025 -0.0189 -0.1072 0.2868 0.4742

86 0.0110 0.3m -0.0319 -0. ml2 0.2868 0.4612

87 0.0030 0.3025 -0.0319 -0.1072 0.2868 0.4532

88 0.0030 0.3025 4.03 19 ~.1104 0.2324 0.3956

97 xm4a 0.1945 -0.0603 -0.1724 0.2604 0.1762

98 -0.0960 0.1915 0.0537 -0.1724 0.2680 0.2448

99 4.0690 0.1915 0.0367 -0.1856 0.2680 0.2416

loo -0.0690 0.2065 0.0447 -0.1520 , 0.2680 0.2982

Removing Doubts Just to prove to myself that testing for single

years using the CAP testing method was more reliable I ran the profitability test for blocks of 5 years each for moving average periods 1 to 100 using the same trading rule.

The results of one of the profitability runs illustrated the point (see Figures 8A and 8B) . This was for the profitability run for the five year period 1986 to 1990.

The table shows that the highest gainers were moving average periods 17 and 16 respectively. Moving average period 13 came in seventh. If I had taken only the total profits as indicators, I would have assumed moving average period 17 to be my best bet. However, I could see from the table that both moving average periods 17 and 16 made losses in 1989. They both failed the CAP test.

This meant that using the results of profitability tests of blocks of five years would not help us improve our annual profits. For Consistent Annual Profits we have to test blocks of one year each.

Therefore, using single year tests has demonstrated that moving average period 13 was definitely the best when used with the moving average crossover rule as applied to British Pound futures. I will use it for my trading next year. I could also use any of the other surviving moving average periods. How- ever, for maximum consistentprofitability the choice has to be moving average period 13.

The Final Say The application of the CAP testing method to

the moving average crossover rule as it was tested here led us to the following conclusions:

l Moving average periods that perform the best in a number of years are not necessarily the most reliable.

l The highest single year gainer is not the best moving average period.

l A high gaining moving average period in any one year may produce a loss in the next year.

l Good performance in bad years does not prove that a moving average period is best for all years.

l The total profit over a period of years does not prove that the moving average period is consistently reliable. On close examination losses were discovered in the individual years tested.

l The moving average crossover rule is a reliable and reasonably stable trading system. All moving average periods made at least some profit in the long run.

l The moving average crossover rule performs best in trending markets and worst in consolidating ones. A reliable moving average period could still ride the bad times and produce a profit or at least cut the losses to a minimum.

l As far as British Pound futures are concerned it seems from this test that the best moving average periods are grouped together from 11 to 14. This only applies to British Pound as it is tested here in one long continuous contract. Other commodities and other trading rules would have to be tested for their own individual findings.

l As with all trading rules continuous monitor- ing of reliability and performance of the system is essential to maximizing profits and minimizing losses.

Conclusion CAP is a testing method, not a trading rule. The above exercise was carried out to illustrate

how to apply the CAP testing method. It in no way attempts to claim that the moving average period 13 is the universal variable to be used in trading. It does not intend to state that the moving average crossover method is the ideal trading method. The findings of CAP as applied above are specific to this case study

What the investigation does imply is that CAP is a method worthy of further investigation. CAP attempts to test a trading rule’s consistency in producing profits. The primary feature is to test trading rules over blocks of trading periods in line with our real life profit reporting periods as apposed to testing over a large historical block. In the above illustration we showed how to analyze the findings of CAP for blocks of one year over a total period spanning seventeen years.

Anyone attempting to use CAP may choose to use a different trading rule, different market or different commodity. The investigator may choose to include his or her own specific trading costs. The close or next day open or for that matter any price may be chosen for the investigation.

With so many different markets and instruments to trade, so many different analytical indicators and trading styles and circumstances there is no limit to number of trading rules and variables that we could use. We are faced with the awesome task of sifting out a trading rule which will make money consistently CAP attempts to find such a rule through

~ rigorous testing. As with everything else in technical analysis

CAP is not infallible. It should be used with care and in conjunction with other analytical tools.

MTA JOURNAL / WINTER 1994 - SPRING 1995 dZii

r BIBLIOGRAPHY

Appel, G. and Hitschler, F, Stock Market Trading Systems, 1980.

Bernstein, J., Timing Signals in the Futures Market, 1992.

Bonini, B. and Hausman, Quantitative Analysis for Business Decisions, 7th Edition, 1986.

Colby,R. U? and Meyers, T. A., The Encyclopedia of Technical Market Indicators, 1988.

Kaufman EJ., New Commodity Trading Systems and Methods, 1987.

LeBeau, C. and Lucas, D., Computer Analysis of the Futures Market, 1992.

Lukac, LX, Brorsen, B.W, and Irwin, S.H., A Comparison of Twelve Trading Systems, 1990.

Pardo, R., Design, Testing and Optimization of Trading Systems, 1992.

Vince, R., Mathematics of Money Management, 1992.

Muneer Al-Hula& worked for nine years as a foreign exchange dealer for banks such as ABNAMRO, Swiss Bank Corp and National Bank ofKuwait in the Middle East. He started getting interested in Technical Analy- sis in 1989. Most of his work involved testing and experimenting using the computer He is an Associate Member of the MTA and a member of Mensa. He has recently moved to Toronto, Canada where he now lives with his family.

4.4 MTA JOURNAL / WINTER 1994 - SPRING 1995

The Klinger Volume Oscillator (WO): A Theoretical Model by Stephen J. Klinger

Introduction The importance of volume analysis and its subsequent application to price movement is widely recognized by technical analysts as requisite to any serious examination of both stocks and market averages. Price itself is the most important dimension of market analysis, but price movement is a function of the intensity of the volume that produced it. The true measure of durability behind price movement is most readily available through volume data. Since volume is a proxy for money flow into and out of a security, volume analysis can expose the internal dynamics (strengths and weaknesses) of price action. Frequently, volume divergences exist behind the cover of price action and provide the only evidence of an impending reversal. Moreover, since volume is believed to precede price, its usefulness as a leading indicator of price strength or weakness can result in a more accurate assessment of price action. Finally, by converting volume statistics into an effective timing model, the technician is better able to gauge the validity of a given price movement.

Background For years, technicians have used volume as a

confirming indicator of price movement. Rising volume is believed to accompany an uptrend, thus confirming the trend. Conversely, a price correction generally occurs on declining volume often capitu- lating on an intense, but short-lived liquidation. When volume deviates from the norm, price action should be called into question. Although volume is subject to distortions such as options expirations and program trading that might tend to skew volume data, its efficacy as a barometer for stocks and market averages has been widely accepted by market analysts.

In the late 1960’s, Joe Granville began looking at volume in a more creative way when he developed a cumulative volume line referred to as “On Balance Volume” (OBV). Granville constructed the OBV by adding the total daily volume to the line on an “Up Day” (Accumulation) and subtracting it on a “Down Day” (Distribution). Although simple in design, it can be quite useful in identifying volume divergences in long term trends of market averages and stocks.

However, because OBV tends to confirm price extremes, it’s usefulness in short term price movements and trading range markets is limited.

Where Granville compared closing price data to determine whether a stock was under accumulation or being distributed, Larry Williams compared the closing price to the opening price. Unlike OE3V Williams added or subtracted a percentage of the daily volume to construct an Accumulation/Distribution line resulting in a more responsive indicator of volume divergences.

Volume analysis took on a whole new dimension with the development of the Chaikin Oscillator. Mark Chaikin expanded on the earlier work of Larry Williams by replacing the “Opening Price” with the “Average Price” (high + low)/2 to construct a volume accumulation line based on the premise that “the higher a stock closed above its midpoint, the more accumulation there was. Conversely, the lower a stock closed below its midpoint, the more distribution there was.” Chaikin converted the data into an oscillator representing the difference between a three day and ten day exponential moving average of the volume accumulation line resulting in more clearly defined buy and sell signals when compared to price action.

In summary, the volume work of Joe Granville and Larry Williams represented a departure from the broad-stroke approach of the past. While “on balance volume” and Williams’ “Accumulation/distribution” are popular volume models, they were developed to identify volume divergences in longer term trends of market averages. However, the Chaikin Oscillator was designed for application to short and intermediate term price movements of stocks and commodities. When used in conjunction with other technical indicators, the results can be quite outstanding.

Rationale Stocks and market averages advance and decline

in irregular patterns that deceive the majority of investors. A stock or market average can be under accumulation while closing down and in distribution while closing up. Although infrequent in dynamic


price trends where the trend is clearly delineated, it Method can be quite common in many of the price patterns In order to calculate the “volume force”, the that occur during a complete price cycle. Moreover, trend (T), daily measurement (DM), and cumulative it may explain prices making higher highs and higher measurement (CM) are required. The trend is de- lows, yet closing down, remaining in an uptrend and rived by summing today’s high + low + close and com- vice versa. Recognition of this phenomenon and its paring it to the previous day’s result based on the inclusion into volume analysis is requisite in the ac- following criteria. curate assessment of price action. 1. When today’s sum is greater than the previ-

While traditional approaches to volume analysis ous day’s, accumulation has occurred and a (1) is do a good job of harnessing raw volume data into assigned as the trend. useful volume models, many unconfirmed trading 2. When today’s sum is less than the previous tops and bottoms still go undetected. The KVO was day’s, distribution has occurred and a (-1) is assigned created to provide technicians with a volume indica- as the trend. tor sensitive enough to signal trading tops and bot- 3. When equality exist (the sums are the same) toms as well as accurately reflect the flow of money the existing trend is maintained. For example, if the into and out of a security. The KVO is based on the previous day’s trend was -1 (distribution) and today’s following tenets. sum is the same, then the previous days trend of -1

1. A price range reveals the limit to which buy- is still in force and vice versa. ers and sellers are willing to purchase and sell a The daily measurement (DM) is the difference security for on any given day While a price range is between the daily price high and low, expressed math- a measure of movement, volume is the impetus be- ematically (high-low) and contains the basic hind that movement. Therefore, the sum of the data requisite to quantify the “volume force”. Finally, high+low+close defines the trend. A stock is under the cumulative measurement (CM) is the cumula- accumulation when the sum is greater than the pre- tive total of daily measurements (DM) in the direc- vious day’s total. Conversely, distribution occurs tion of the trend originating and terminating on when the sum is less than the previous day’s total. trend changes. For example, if the previous day’s When equality occurs, the existing trend is main- trend was -1 (distribution) and today’s trend is 1 tamed. By redefining volume accumulation/distribu- (accumulation), the cumulative measurement (CM) tion in this fashion, the KVO incorporates the phe- is the sum of today’s daily measurement (DM) and nomenon mentioned earlier into its analysis. the previous day’s daily measurement (DM) origi-

2. Volume produces continuous intra-day changes nating on the trend change. The cumulative mea- in price reflecting buying and selling pressure. Where surement (CM) increases by the daily measurement existing systems focus on one side of the volume equa- (DM) for each consecutive day that the trend is in tion in their analysis, the KVO quantifies the differ- force. Upon termination of the trend, the process ence between the number of shares being accumu- begins again. lated and distributed each day in a security referred The following spread sheet on synoptics commu- to as the “volume force”. It is based on the belief nication serves to illustrate the procedures explained that the “volume force” is the true measure of vol- above. Taking the period 10/19/93, the trend is -1 ume fueling prices towards higher or lower levels (distribution) because the sum of the price range and can therefore improve the reliability of buy and (high+low+close) is less than the previous day’s to- sell signals. Since price and volume normally rise tal; the daily measurement (DM) is 1.50 derived by and fall together, a vigorous and rising “volume subtracting the high-low; and the cumulative mea- force” should accompany an uptrend and contract surement (CM) is 2.75 derived by the sum of today’s over time followed by a rising “volume force” reflect- daily measurement (DM) and the previous day’s daily ing some accumulation before a trading bottom measurement. Note in the example provided how the develops. calculation for the cumulative measurement (CM)

3. By converting the “volume force” into an originated on 10/19/93 on a trend change, terminated oscillator representing the difference between a 34 on 10/20/93, and began again on 10/21/93 on another day and 55 day exponential moving average with a trend change. 13 day trigger (the numerical point at which the The formula for the “volume force” is actually oscillator crosses through a 13 day EMA of the oscil- two formulas combined. It was designed to utilize lator), the technician can track the force of volume the basic data contained in a price range, daily and into and out of a security. Comparing this force to over time, to quantify the difference between the price action can help time buy and sell decisions number of shares being accumulated and distributed by identifying volume divergences that exist at tops each day in a security so that the difference or “vol- and bottoms. ume force” can be isolated and converted into an


Security: Synoptics Communication

DATS 09/24/93 09/27/93 09/2a/93 09/29/93 09/30/93 10/01/93 10/04/93 lo/as/93 10/06/93 10/07/93 lo/oa/93 m/u/93 10/:2/93 x0/13/93 m/14/93 lO/lS/93 m/la/93 10/19/93 10/20/93 10/21/93 10/22/93 10/23/93 10/26/93 m/27/93 10/28/93 10/29/93

HIGH LOW CLOSE VOL. TREND DM CM VF 26.250 2s.soo. 26.SOO 26.000 2S.790 24.500 23.7so 23.SbO 23.2SO 22.2so 22.2so 24.000 2s.000 27.710 27.600 27.SOO 28.000 27.7SO 27.000 27.SOO ta.000 26.740 26.SOO 27.2sO ta.000 27.7SO

24.SOO 23.soo 24.12s 24.000 24.2SO 23.SOO 22.250 22.soo 21.7SO 20.240 21.000 2l.250 24.000 f6.2SO 26.710 26.500 26.750 26.250 ts.710 2s.soo 26.000 2s.soo If.750 26.000 26.7SO 27.000

2S.62S 24.2% 2S.2SO 2s.000 24.100 23.100 22.soo 23.000 22.000 u.soo 21.soo 23.625 24.100 27.2So 27.000 27.000 27.975 26.SOO 26.2110 27.375 26.l2S 26.12J 26.371 27.000 27.000 27.7sO

SO227 -1.000 1.750 2.100 -2009079.750 sla97 -1.000 2.006 4.500 -576633.166 44641 1.000 2.37S 4.37s 382637.500 35347 -1.000 2.000 4.37s -302974.219 22737 -1.000 1.500 21.87s -1112661.750 2s904 -1.000 1,000 6.075 -1836628.675 31591 -1.000 LSOO 1.37s -2027402.000 19036 1.000 1.000 2.soo 360759.969 lolad -1.000 LSOO 2.soo -607720.312 12760 -1.000 2.000 4.soo -586222.250 3942s 1.000 1.2so 3.2So 909007.600 29774 1.0000 2.710 6.000 246116.594 2903.s 1.000 1.000 7.000 1129643.000 92747 1.000 1.100 a.500 6001276.500 16991 1.000 0.710 9.2so 1591137.a75 16518. -1.000 1.000 1.7so -231971.464 19030 1.000 1.250 2.250 211444.S31 26636 -1.000 LSOO 2.7So -1a9416.156

9652 -1.000 1.2So 4.000 -361950.000 looao 1.000 2.000 3.2SO 232615.436 14651 -1.000 2.000 4.000 -0.000 12267 -1.000 1.250 s.tfo -642557.062

6767 1.000 0.7so 2.000 .2l9675.000 7911 1.000 1.2So 3.2So. 182561.S31

lrasl 1.000 1.2so 4.500 661377.750 7189 1.000 0.7110 s.2so 513500.000

oscillator. The volume force is expressed mathematically in the following formula:

VF=Vx /(2(DM/CM)-l)I xTx100

where: V= daily volume T= trend DM= daily measurement CM= cumulative measurement

The equation ) B(DM/CM)-1) 1 was chosen because of its property to state a positive result between zero and one. This tendency towards the number 1 is a key element of the equation because the data is being converted into an oscillator. Since the trend (T= 1 or T=-1) delineates whether the “volume force” represents accumulation (1) or distribution (-l), the equation must be expressed as “the absolute value of”. For example, the period 10/19/93 from the previous spread sheet on synoptics communication produced a daily measurement (DM) of 1.50 and a cumulative measurement (CM) of 2.75. Therefore, 2(1.5/2.75)-l equals .09, a positive result. However, it can also produce a negative result as it does with the data from the period 9129193 where 2(2/4.375)-l =-0857. Applying the data from 10/19/93 to the entire formulawhere (DM)=1.50, (CM)=2.75, (V)=20836, and (T)=-1, the “volume force” can be computed.

20836 x 1(2(1.50/2.75) -1)l x-lx 100 = -189,418.156.

The weighted result of -189,418.156 is the “volume force” for the period 10/19/93.

The KVO is a momentum oscillator of the volume force. Mathematically, it is expressed in the following formula:

KVO = MOV (VF( ),34,E) - MOV (VF( ),55,E)

where: 1. MOV(VF( ),34,E) is a 34 day exponential

moving average of the volume force. 2. MOV(VF( ),55,E) is a 55 day exponential

moving average of the volume force. 3. The formula to calculate an exponential

moving average (EMA) = (C x A) + (E x B), where C = today’s volume force A+2/ (X+1), where “X”=movingaverageperiod E=yesterday’s EMA B=l-A

Results The KVO was tested using Me&stock software

and two 400 day test periods consisting of the high, low, close, and volume. Two passes were conducted on each security spanning the period g/30/91-11/25/94. The securities selected for the study were:

1. American Bar-rick (ABX) 2. EMC Corporation (EMC) 3. International Business Machines (IBM) 4. Dell Computer (DELL) 5. Intel Corporation (INTC)

J


Illustration of the KVO on Dell Computer for 520 Trading Days

-200000-

I

L

1s0000

100000

soooo

0.00

-10000

- LOOOO

- isooa

-2ooot

50.00

w.00

90.00

3s .oo

i

20.00

ss.00

10.00 u

‘22 It?,OCT HOUOCC JIIW CLIMR Mlt IIWJUM JULMII SLI OCT HOU DLC JRN rC8tlM IIPRHAV JUtl JULIUfi

so.00

9s .oo

90.00

JS .oo

BO.00

21.00

20 .oo

is .oo

to .oo


Summary of Results for KVO Period: 9/30/91- U/25/94

PASS #1:9/30/91 - 4/28/93

PASS #2:4/29/93 - 11/25/94

The following conditions were preset into the profitability system to insure the validity of the test results.

Buy Signal This system generates a buy signal when the

KVO crosses above its 13 day exponential moving average and the security closes above its 34 day exponential moving average.

Sell Signal This system generates a sell signal when the

KVO crosses below its 13 day exponential moving average and the security closes below its 34 day exponential moving average.

summary A common technique used by traders with oscil-

lators is to initiate long or short positions on the crossovers above and below the centerline. While this is a popular method, it can limit the effective use of the KVO. The dominant characteristic of the KVO is the velocity at which the “volume force” precedes price alerting the technician to short-term price extremes. With the use of a trigger, trades can be initiated prior to the crossover of the zero line and a reversal in price.

The most powerful use of the KVO is registered when the indicator reaches an extreme reading either above or below the centerline and reverses direction. This signals an exhaustion of the prevailing price trend and warns of an impending price reversal. Although the KVO does an excellent job of interpreting the volume force behind price action, it is not a panacea. The test results reflected the general KVO

methodology but no attempt was made to maximize the effect by using the aforementioned “powerful use”. One result of this was that there were approximately the same number of profitable trades as unprofitable ones. Never-the-less, the methodology did identify the significant moves resulting in overall meaningful profits.

While the KVO works well in timing trades in the direction of the trend, it is less effective against the trend. This can create problems for the trader trying to scalp a trade against the prevailing trend. However, when the KVO is used in conjunction with other technical indicators better results can be achieved. I recommend using Williams %R for confirming an overbought/oversold price condition and Gerald Appel’s MACD to confirm the short-term direction of price.

For the best signals, the following guidelines should be followed.

1. The most reliable signals occur in the direction of the prevailing trend. Trades against the prevailing trend tend to be high risk ventures. Strict stop guidelines (i.e. failure to penetrate the zero line and a violation of the trigger) should remain in force.

2. The most important signal occurs when the KY0 diverges from price action especially on throw-over’s (new high) or throw-under’s (new low) in overbought/ oversold territory. For example, when a stock makes a new high or low for a cycle and the KVO fails to confirm this, the oscillator is warning that the trend is losing momentum and nearing completion.

3. A stock in an uptrend (>89 day EMA) should be accumulated when the KVO drops to unusually low levels below zero, turns up, and crosses its 13 day EMA trigger. Conversely, a stock in a downtrend


1

il I E !9999S555??? f ?4Ssa;EEtEE


, ,115 lucu“ .4‘uum , 11-a ‘us,,u m4nmhm 8‘l-- aar‘uwm D WI wmm 64wml‘a i ‘I- w/It& #4umm

W“ ,11- ‘&mm sm-9‘ 8 4cm?Jz‘ sol%‘ .1*- ‘nmm p’1‘ @ 4w‘om UC-.‘ , “W csmll‘ Ll’l‘ , 4 cmw11 1‘1’,‘ . 11-l Nnul‘ ~*,I . -4 cuw11 ml-m a ‘,=s wsm ac’n 8 4cuwmt ““1‘ 8 ‘I- - 8Icvlo4wnm

S-II-8 - -WI* - WWY -‘.u -* “-a

-,t-V-0 -*- -‘rut---i

‘uw11=‘-

(~89 day EMA) should be sold or shorted when the KVO rises to unusually high levels above zero, turns down, and crosses its 13 day EMA trigger.

In conclusion, the KVO does an excellent job of harnessing raw volume data into an effective timing model. While more research is needed on volume analysis, its efficacy is well established in the field of technical analysis. The KVO has demonstrated the potential to make a contribution to the overall body of research on volume analysis.

BIBLIOGRAPHY I

Beckman, R.C., Supertiming, Los Angeles, California: Library of Investment Study, 1979.

Colby, R.W. and Meyers, T.A., The Encyclopedia of Technical Market Indicators. Homewood, Illinois: Business One Irwin, 1988.

Murphy, J.J., Technical Analysis of the Futures Markets. New York: New York Institute of Finance, 1986.

Manual-Metastock Version 2.50. Salt Lake City, Utah: Equis International, 1988-1991.

ACKNOWLEDGEMENTS I

I would like to thank Richard Elliott Dysart of Trendway Advisory Service, Louisville, Kentucky for his patience and willingness to teach me the principles behind his theory on price projections and whose technical work has inspired me to develop the KVO. I would also like to thank Dr. Lothar A. Dohse, Chairperson, Department of Mathematics, University of North Carolina at Asheville for his time and suggestions in reviewing my formulas and Walter G. Murphy, Jr., senior international market specialist at Merrill Lynch for encouraging me to pursue the CMT designation and for always finding the time to answer questions and educate me in technical analysis over the past six years. Finally, I would like to thank Marlin Newell, president, Mid-Atlantic Business Systems and lifetime friend, for his expertise in computer programming and assistance in writing the program for the KVO.

Stephen J Klinger has worked in the brokerage business as a Financial Consultant since 1985. He is currently a Senior Financial Consultant with Interstate/Johnson Lane, headquartered in Charlotte, North Carolina. Mr. Klinger holds degrees in Business and a Masters in Education. He is a level III CMT candidate and an applicant for membership in the Market Technicians Association.


Improving Returns While Controlling Risk: Integrating Wyckoff’s Tools with CANSLIM Stocks by J.C. Coppola, III

Viewpoint found that the underlying factors and common char- It is my contention that little has changed in the acteristic of the most successful investments tended speculative nature of campaigns that take place to be pretty much the same, stock by stock, cycle after on Wall Street. The simple reason for this is that cycle. Another interesting point gleaned from O’Neil’s human nature runs relatively constant through the study was that “the vast majority of the price boom and bust cycle. In the financial markets, a increases began emerging out of areas of a price con- collective human bias renders a decision. This de- solidation pattern that had occurred over several cision takes the form of price on an electronic quote months.” machine, or at the post on the floor of a global Just as William O’Neil had modeled the charac- exchange. Price influences other participants and teristics of the greatest winning stocks, 1953-1983, then they make decisions according to prevailing Richard D. Wyckoff had modeled the campaigns of market trends. Price, therefore, plays a profound some of the greatest stock market operators, X389- role in shaping the future outcome for many 1928. Wyckoff also found many common character- events, be it in global stock markets, currency mar- istics among successful investment campaigns. After kets, or international credit and regulatory cycles. modeling the action of Jesse Livermore, Edward What academia and economic theory seems to miss Wasserman, James Keen, J.l? Morgan and many is that price is always the residual of the collective other big operators of his day, Wyckoff developed a bias presented by the financial community This paradigm which helped to explain the boom and bust human sentiment, and the supply and demand that cycle in stocks. He implemented this model and grew it creates, is what an investor applying technical his account such that he eventually owned a man- analysis strives to exploit in his goal to make sion next door to the Alfred Sloan Estate in the money in the financial markets. Hamptons.

“Market participants base their decisions not on objective William O’Neil has developed an excellent sys-

conditions but on their interpretation of those conditions. tern that lends itself extremely well to computerized

When this process called ‘reflexivity’ takes place, the hu- stock screening. In overlaying Richard Wyckoff’s man bias becomes one of the fundamentals which shape the model of an idealized trading range, it becomes evolution of price.” apparent that we have increased insight into the

George Sores, The Alchemy of Finance timing of our buy and sell decisions. The Wyckoff method specifically aids us, stock by stock, in these

Introduction “areas of a price consolidation pattern” that occur In the following research, we supplemented over several months prior to a historic mark-up.

William O’Neil’s approach to growth stock invest- Hence, a very nice correspondence between the two ing by applying tools of technical analysis that methodologies is born. Richard D. Wyckoff developed circa 1920. Our study In conducting our study of CANSLIM and is not trying to prove a causal effect or strict linkage Wyckoff, we critically surveyed a database of 78 Wil- between O’Neil and Wyckoff, but rather we are try- liam O’Neil CANSLIM stocks, applying Wyckoff’s ing to provide the investor with some practical tools methods of technical analysis to those charts. We were which are associated with the Wyckoff method of attempting to ascertain if Wyckoff s model was con- technical analysis. Our work will show you that Wyckoff s tools proved very helpful in timing deci-

gruent and helpful with the same stocks O’Neil modeled in his study

sions in many of the stocks that William O’Neil mod- In the following paper we will describe what eled in his study, The Model Book of the Greatest was learned from CANSLIM and how the applica- Stock Market Winners. tion of Wyckoff s system of technical analysis can

O’Neil’s study modeled 515 of the best perform- aid the investor in improving returns and control- inglisted stocks from 1953-1983. Fortunately, O’Neil ling risk.

I


What Did We Learn From William O’Neil? In his book, How to Make Money in Stocks, O’Neil

introduced what he learned from his study The Model Book of the Greatest Stock Market Winners to the general public. In this book he coined the acronym, “CANSLIM” to summarize the characteristics of winning stocks. The characteristics that CANSLIM addresses are defined below:

CANSLIM CHARACTERISTICS

The Basis for Stock Selection

C-Current quarterly earnings per share. A company should have large increases in current quarterly earnings per share. These are usually greater than 25% and/or have a recent acceleration in the rate of quarterly earnings per share.

A-Ann& earnings increases. The annual earnings per share should show consistent growth over the past five years. This can be defined as a 5-year growth rate greater than 25%.

N-Stands for new highs. Buy when a stock is hit- ting new highs in price. Always be on the look out for companies with new products, new service, new management, or major improvement in industry conditions.

S-Supply and demand. The number of shares outstanding of the stocks under consideration should be under 30 million, preferably around five million.

I.-Leading stocks in leading industries. Buying should be concentrated in leading stocks in one of the strongest groups.

I-Institutional ownership. You want at least one institution owning the stock you are considering buying.

M-Stands for market direction. Analyze the market. Is the Dow Jones Industrial Average above or below its 200-day moving average? How long has the 200-day moving average been in an up trend or downtrend? Analyze the daily Dow’s price and volume movement every day Equally as important, how are your individual stocks performing in the market environment.

Source: William O’Neil + Co.

After surveying the O’Neil basis for stock selection it becomes apparent that it is comprehensive. The first six characteristics of CANSLIM are directly attributable to the stock. These characteristics are consistently present in areas of price consolidation

that occur over several months prior to the mark-up phase beginning.

C - current earnings A - five year earnings growth N - new price highs s- shares outstanding L - leading stocks in leading industries I - institutional ownership, and M - market is studied independent of the stock

Interestingly, the “CANSLIM” variables lend themselves well to computerized searching. There- fore, after constructing a model based on these fundamental characteristics, we then need a technical system with the anchors necessary to filter out the many stocks that come up. This system must tell us precisely when to take a position and how to control our risk.

What is Technical Analysis and Why Wyckoff? Technical analysis is the study of supply/demand

and psychological factors which aid a speculator in his/her quest to anticipate future price moves of individual stocks, sectors of the market and the stock market itself A strong technical system empowers the investor with the keen power of observation, it is based upon sound principals that manifest stock market behavior and human nature, while always controlling risk. Therefore, a disciplined model in technical analysis should aid the investor immensely when implementing a trading philosophy which applies the CANSLIM computer searching potential.

As we all know, owning shares of a company is not the same as owning the company itself Clearly, shares are bought and sold each day as a result of portfolio positioning, sector weighing, excessive cash reserve build up. Additionally, there may be a need for cash for other investments, or simply a manager may react to current stock market conditions. Also, each investor has a different perspective, time-frame and objective for the investment. Warren Buffet’s methodology and time-frame may differ from Richard Driehaus, which could differ from Marty Schwartz which may differ from a specialist who has a position in a stock at any given moment in time. Nonetheless, each player is investing to maximize his or her util- ity. The players share the common goal of making money, but on any given day, the reason for buying or selling their shares may differ. Hence, supply and demand are always influencing share prices.

A good technical system could be considered the most integral part in the investment equation. It tells us when to actually make an investment and how to control our risk. Richard D. Wyckoff’s system is very definitive. It is a system that is intermediate-term


in nature (3-18 months). It is built for those who try to buy long, or sell short, as close to the extremes in the trading range as possible. It is specific in controlling risk through the strict use of stop limits. Wyckoff money management overlay is guided by the principal of pyramiding one’s position by scaling in one-third, one-third, one-third, while trading in harmony with the larger trend that is developing.

Wyckoff’s Solutions CANSLIM’s check list is excellent, however, it

can be somewhat ambiguous and leaves an investor doubtful and hesitant to take action in a real time trading environment. Mr. Wyckoff s trading range model helps us refine and solidify the efforts of O’Neil.

The three concepts of Wyckoff that specifically add significant value in timing are:

1. The Idealized Cycle and Primary Market Phases 2. The Composite Operator 3. The Nine Buying Long and Selling Short Tests

to CANSLIM is the concept of the Idealized Cycle and the Primary Market Phases (see Figure 1). This simple but effective tool serves as a basic starting point in our analysis of a stock’s pricing cycle. Spe- cifically, it helps us answer, “Where are we in the pricing cycle?” This concept was fully described in the Market Technicians Association Journal article by James Forte, entitled “Anatomy of a Trading Range.”

The Composite Operator Wyckoff was thorough in his analysis of the trad-

ing range. Another invaluable tool that Wyckoff provides is the concept of the “Composite Operator.” Simply, Wyckoff felt that an experienced judge of the market should regard the whole story that appears on the tape as though it were the expression of a single mind. He felt that it was an important psychological and tactical advantage to stay in harmony with this omnipotent player. By striving to follow his foot prints, Wyckoff felt we are better prepared to grow our portfolios and net-worth.

These techniques enforce discipline in the pro- By implementing the concept of the Composite cess and force the investor to be systematic and Operator, the investor is emulating positive trading consistent with his analysis and trading. behavior. Through Wyckoff’s rigorous test of volume

and price action, we are more in tune when follow- Wyckoffs Idealized Cycle and Phases ing the clues of the tape while viewing them through

The first area of Wyckoff that offers value added the eyes of the Composite Operator. The following is

FIGURE 1

Idealized Cycle

olsmtnunow AIIEA

Conception of PrMary Market Phases

DlSTRISUTtON

CAUSED II GllEtD ACCUMULATION

Aaumulatlon: The cstabllstunent of an lnvcstmcnt or spauJaUvc posltlon by prolcss~onal lntcruta ln anUclpaUon of an advana In price.

CAUSED I” FEAR

Markup: A sustahcd upward prkc movement.

DLtrlbuthn: the clhhatlon of a long hvcstmcnt or rpcculatJvc pod&m.

Markdown: A sushlncd downward price movemcnl.

Source: James Forte, MTA Journal, Summer-Fall, 1994.


L

FIGURE 2

.._

a verbal and graphical depiction of the entire cycle the Composite Operator follows from the distribution to accumulation phase.

The Campaign As “Bid” and “Ask” spreads jump, and the differ-

ential between open and closing prices or “spread” widens positively, the Composite Operator is marking the market up, luring and tempting the less sophisticated players who are attracted by rising prices. At this stage, in the mark-up phase, demand has control over supply. Prices move easily on fairly low volume. Prices continue higher until the move is abruptly halted. Preliminary Supply (PSY) is reached, as the Composite Operator begins to unload a very heavy line of stock on the market. (See F’igure 2. )

his size in a single session, for it takes several weeks or often months. Therefore, this point of preliminary supply is only the beginning of the distributive process. Subsequently, the keen investor will begin to notice that the stock no longer moves easily nor has the vitality it once did. The Composite Operator is unloading on the way up. His supply dulls the strength of the market.

Heavy supply finally breaks the market as the Composite Operator throws his final line at the market’s top. A Buying Climax (BC) results as prices fall from their highs on the day/week and “spread” significantly widens. The Composite Operator sells on the way down as prices collapse violently and creating an Automatic Reaction (AR) to the Buying Climax (BC) results. (See Figure 3.)

Nonetheless, the big fellow could never work off The stock becomes heavy, reacting in dull fashion

FIGURE 3


FIGURE 4

.................................................................................................................................................

.... w ................................. ..n........................~ .................. a.*

...... ................... ... .... .................

,,: A+ .::

...:.:::.::.::::::::::: f~:‘; :::::: ..

.............

+

::::::::Q

........... tr

;;+.fl ::::.:::::::.

.tT..

;:;

..............

..........

.......

:’ ......... ............... :e-y+~::.:jlr’~‘l~.~ I: .. lj:::::~~~;;,

............

11. ............ t..

..t?:::: :::

................................................ .... x.o

....... .............. .........

+t:.:.:::.::.:::.: .I

.............. il. ..................................................

......... $.I

=.,

.. .T.. ... . . ........ T.. .............. ........................................... .. .

.::::

:&I

r::. :‘.‘IT :...:.....:.::.::.:..::..:.:::.::.:::.:.:.::! ..::.:: :::. ::::::.:::::::::.::.::.::::::::::::::::::::::::: :::::::::::.::.

a.o

.. El:

.............. ............ ............ ............................................................................... b

=., 1 ..................................... .... .......... ............. .............................................. ................................ ..

................... ......... ....... ........... .... ....... ................. ................... ............ ..... .......... p.*

.......................................... ....................................................................................................... ~.D

........................... ......... .... .... ............. ................. 2,.*

~~~.~t~l.;;i.~l~;l.;..t~ 11.0

.... ~l~.i.~..~.~.~;.;.;.~.i..t.;.~~~.t~i.;.~i~~..i~;~.~.~~.~.~.~.;~i..~..~~~ .... mms I III,

‘71 ca I* 12 za J cm II 1s a J 10 17 z-l .

FIGURE 5 ._. .-

FIGURE 6


and even perversely to seemingly good news. The upward momentum slows as the public’s buying effort shows little result. A Trading Range (TR) forms, and towards the top of it, the Composite Operator heavily supplies the market whenever large bids are available. Near the bottom, he holds the market and creates minor support. However, throughout this range, he is always a net seller of stock. He is in the final phase of accomplishing his upside objective and working off his position. As each week passes, he slowly distributes his entire line of stock. The stock seems heavier and heavier and his work in this range is almost done. The Composite Operator’s net position is now zero. Nevertheless, this omnipotent operator will soon switch his gears. He begins using rallies to the top-end of the trading range to sell stock short. Prices swing as lower tops and lower supports are often experienced within the trading range. The stock runs up and tests the highs a second time. Wyckoff calls these a series of Secondary Tests (ST). Supply continues to overwhelm demand at the top the Trading Range (TR) as the stock makes little or no net progress to the upside. (See Figure 4.)

The Composite Operator sometimes even rallies the stock hard through the top of the trading range. This will be one last Upthrust After Distribution (UTAD) before prices fail quickly and the stock plunges back into the trading range. This final move tends to purge the weak short sellers and instills hope in the minds of the public who are long in the stock. Prices eventually fail, as the differential between open and closing prices or “spread” widens and the old upward stride/trend line is broken. Positive momentum is now clearly gone and a minor downtrend within the Trading Range (TR) is established.

, (See Figure 5.)

Supply quickly overwhelms demand and prices fall violently, thus piercing the bottom of the trading range. A Sign of Weakness (SOW) is experienced. When the market attempts to rally back, it runs into supply near the bottom of the old trading range. A Last Point of Supply (LPSY) results. Unfortunately, for those who are long in the stock, the highs have already been experienced. The SOW signifies important weakness and the barrier at the LPSY turns to ICE. This line of ICE, where old support becomes new resistance, will not be significantly penetrated by the bullish public until the Composite Operator is finished with his short sale campaign. Because the big fellow is no longer there to support the market, supply begins to control the market. The stock seems to sinks of its own weight. (See Figure 6.)

The market continues its downtrend with “spreads” widening as prices are forced lower. (In severe mark-down phases, prices accelerate climac- tically as demoralization sets in.) At this point, it is clear that the stock will not see its old highs, soon. A disheartened public quickly rushes to unload the last of their holdings. Suddenly, substantial buying begins to provide pronounced support, Preliminary Support (PS) is experienced. (See Figure 7.)

Volume expands, spreads widen and prices begin to close stronger near the top for the day The stock follows through and the public sees this strength as the last opportunity to liquidate. The Composite Operator covers his short position throughout, forc- ing prices higher. It is too early to buy, but he has accomplished his downside objective. The stock falls back to the low ground and climaxes as the Compos- ite Operator covers the last of his line. High volume is apparent and the down move is stopped, as a Sell- ing Climax (SC) is reached. The market reacts and

FIGURE 7


FIGURE 8

I CM LTO.

an Automatic Rally (AR) ensues. The Composite Op- erator’s final short covering and initial buying forces the market upward quite easily This results in the downward stride being broken. The Composite Op- erator lulls the market and works out all weak players as he accumulates his line near the bottom. Prices then fail for there is a lack of a concerted demand. At this point, short covering and minor demand are driving the market. The market undulates in a range as the stock backs and fills, testing for large sellers.

Throughout the early stages of this trading range, the Selling Climax (SC) area is again Tested (ST). The stock is almost dormant, trading dull and qui- etly for several sessions (sometimes weeks to months). As the stock trades in this tightening range, with the “spread” narrowing, the Trading Range (TR) continues to form. The floor traders get bored with the stock as the public loses interest and volume shrinks. Spreads narrow at the bottom of the Trading Range (TR) and the market begins to coil, tighter and tighter. (See Figure 8. )

Minor supports and higher tops form within the trading range as the Composite Operator continues his quiet buying. Several weeks (often months) pass since the old highs were reached. The Composite Operator’s line is almost complete and the stock is prepared for its mark-up. He attempts to push the stock lower, but notices little stock offered. In fact, the bids are willing as the stock is tight and snaps back quickly The longer the trading range exists, the more likely he will violate its bottom to purge the final weak holders. On each thrust to the bottom of the range, it becomes clearer to the big fellow that demand has control. Eventually, his buying becomes aggressive enough to force prices up very easily (See Figure 9.)

Prices spring from this point and, often times, move upward quickly Again, the stock pulls back but little stock is offered to the market. A Last Point of Support (LPS) is formed and a powerful thrust in volume occurs as a Sign of Strength (SOS) results. Another LPS forms on tight spread with volume contracting significantly. Price “spreads” are tight for several sessions until an enormous surge in volume takes place. This represents the Jump Across the Creek (JUMP) of floating supply Every offer is hit at the market as prices move easily and another Sign of Strength (SOS) is experienced. A reaction or Back Up to the Edge of the Creek (BAC) then forms on shal- low volume as demand takes control. (See Figure 10.)

The floor brokers and speculators notice the interest. Psychology begins to turn, news begins to look brighter and the sun begins to rise once again. Specu- lators trade on the way up causing interim reactions as the Composite Operator holds his line and forces prices steadily higher. The stock is in its mark-up phase, pausing intermittently, to re-accumulate as speculators take profits all the way up. Eventually, the public gets interested and psychology begins to turn more positive. Several months and quarters pass. Euphoria sets in. And, everybody must own the stock! Once again, the Composite Operator successfully played his campaign, round trip. (See Figure 11.)

Nine Tests In conjunction with accumulation and distribu-

tion, and viewing the market through the eyes of the Composite Operator, Wyckoff affords nine specific tests to apply to all trading ranges. These tests mesh well with CANSLIM and on the next page we list them. It’s not the purpose of this article to teach the


FIGURE 9 - cm.

FIGURE 10

FIGURE 11 nsa LTD. o&2+/7*


reader the technique of Wyckoff. He wrote many books that are still available through Fraser Publish- ing. Also, the Wyckoff/Stock Market Institute in Ari- zona and Golden Gate University, in California offer extensive course material based on his methodology As a synopsis, here are the nine buying long and selling short tests. Later we will apply them to O’Neil’s model studies to show you Wyckoffs tools in action.

Wyckoff developed nine tests to apply to all trading ranges prior to taking a position. By applying these tests, the investor is more consistent and disciplined when analyzing potential long and short candidates. The nine tests were specifically developed to control risk. (See Figures 12A and 12B.)

Figure 13 is a chart of Computer Associates. On this chart we have applied the Nine Buying Long

Tests to an area of price consolidation prior to a major period of mark-up.

Methodology of the Study William O’Neil + Co. provided us with the can-

didates, 1970-1983. Of these 273 stocks, we collected data for every stock which we could retrieve one year of reliable data prior to where O’Neil’s idealized model would suggest that one could have bought, and one year after O’Neil’s sell model could have signaled. Our study focused on the area where stocks emerged “out of areas of a price consolidation pattern that had occurred over several months.” It was our belief that the annotated trading range that Wyckoff developed would fit nicely with this basing pattern that O’Neil evidenced.

FIGURE 12A

3a.) NINE BUYING LONG TESTS:

INDICATION:

I .) Downside objective accomplished

DETERMINED FROM:

2.) Activity Bullish (volume inueases on rallies and decreases on reactions)

3.) Preliminary Support & Selling Climax

4.) Average or stock stronga than market (i.e. more rqmnsive on rallies and more resistant to reactions)

5.) Downward Stride Broken (i.e. Supply line penetrated)

6.) Higher supports (daily lows)

7.) Higher Tops (daily high prices rising)

8.) Base forming (boriwntal price line)

9.) F&mated profit is at least three times the indicated risk (Vertical Chart for stop order placement)

Vertical and Figure Chart

Compare Vertical Chart

Vdcal or Figure Chart

Vertical or Figure Chart

Vertical or Figure Chart

Figure Chart for profit objective

Pichard D. Wyc!df

FIGURE 12B

3b.) NINE SELLING SHORT TESTS:

INDICATION:

I .) Upsi& objective accomplished

2.) A&&y Bearish (volume decreases on rallies and incxeascs on rca&onr)

3.) F’reliminq Support & Buying Climax

4.) Average or stock weaker than market (i.e. more rsponsive on reactions and sluggish on rallies)

5.) Upward Stride Broken (i.e. Support line peneeated)

6.) Lower supports (daily low prices falling)

7.) Lowe Tops (daily high prices falling)

8.) Craw forming (lateral movement)

9.) Estimated profit is at least three times the indicated risk (Vertical Chart for stop order placement)

DETERMINED FROM:

Figure Chti

vcttical chart

Vcttical and Figure chart

Compare Vertical Chart

Vertical OT Figure Chart

Vertical or Figure chart

Vettical (x Figure Cbti

Figure C&an

Figure chart for profit objective

- Sauce &chard D Wyckoff


: f :

: 5 :

: : :

: : :

. ’ : :

:

. :

:

:

: : ,

: :

: : :

: : : I I *

: : :

. : : : . 5 :


The data search resulted in 78 stocks which qualified. Each stock was then adjusted for splits. We utilized the Financial Information Inc. Finan- cial Stock Guide for all accurate history on split ad- justments. After returning the data on the 78 stocks back to its original form, pre-split, we then printed daily bar charts and 1 x 1 Point and Figure charts. Where necessary (i.e. lower priced stocks), smaller point and figure box sizes were used (1/2 x ‘/2; ‘/4 x 1/4>.

It must also be mentioned that during the time span sampled, 1970-83, we experienced 4 bull market phases, 3 bear market phases, as well as 3 trading range markets (one very prolonged). Thus, a diverse amount of market conditions were reflected in the time-frame of the charts studied. Addition- ally, it was our belief that these 78 charts reflected a random sample of the 273 stocks O’Neil utilized in the updated study, 1970-1983.

In the interest of brevity and for the MTA Jour- nal a sample set of charts which represent the relationship that Wyckoff can have with CANSLIM stocks can be found in the appendix of the paper. If the reader is interested in surveying the entire study, it can be made available through the Golden Gate University library or the Market Technicians Asso- ciation library (See J.C. Coppola III, Harmonizing Wyckoff and Canslim.)

Results & Observation of the Study In order to score Wyckoff’s Applicability, we

weighted the Yes results to get a better idea of the degree of importance that Wyckoff Analysis had in the process. + Yes means extremely helpful in the long-term, 6-18 month, trading range that preceded the historic move, = Yes means very helpful in the 3-6 month base that formed prior to the move, and - Yes means marginally helpful in the minor consolidation that took place, a No response means that no consolidation took place and Wyckoff would have been of no value. Below are the results of the study

I TABLE A

The second characteristic looked at was, did Wyckoff’s model for the ideal trading range allow an investor to take a position? Specifically, “Did Wyckoff buy at a lower risk point in the trading range?”

What we learned was that in thirteen instances Wyckoff allowed a position, but at a higher risk point. However, in fifty-five instances, or seventy percent of the time, Wyckoff bought a CANSLIM idea at a lower risk point than CANSLIM.

To reconcile the past two questions, we turned to performance enhancement. How did the addition of Wyckoff s model aid performance? The statistics show that Wyckoff increased performance materially In fact, 2,213% in total or 31.17% per position was added.

TABLE C

We want to reiterate that both the O’Neil and Wyckoff models are paradigms. Both O’Neil and Wyckoff inspected past winners and looked for shared characteristics within these winners. The database of 273 stocks from 1970-1983 represents about 103,581% of performance. This turns a $1 investment, compoundingannually to $1,114,544,804.97 in three years. Now, we all know that this performance is not replicable in a portfolio, however because we are working on the high speed inter- section principal. This principal states that, “if you play in traffic you are much more likely to get hit by an on coming car.”

In our final analysis, the Wyckoff method of technical analysis, when applied to a database of CANSLIM stocks, produced:

1. An extremely congruent application and methodology

2. 84% of the time Wyckoff allowed us to take a

Because the vast majority of CANSLIM stocks tend to emerge out of areas of a price consolidation that lasts several months, the results show that in 71 of the 78 charts Wyckoff proved very helpful.

I TABLE B I

position at the same or a lower risk point (70% actually lower), and

3. Wyckoff analysis when combined with CANSLIM materially aided performance.

For further descriptive value we have developed a legend for Wyckoff terms used during accumulation and distribution and an appendix of selected charts from our study The Appendix can be found at the end of the article.

Summary and Conclusion O’Neil’s CANSLIM approach lends itself well to

computerized searching and filtering techniques to pinpoint the initial characteristics of a winning stock.


APPENDIX

Legend for Wyckoff Terms Used During Accumulation and Distribution

The best way to understand the concept of accun Below we have provided the reader with the model I I

ACCUMULATION SCHEXWATIC

1. (PS) Preliminary Support: substantial buying interest begins to provide support to stop the down move. Volume and spread widen and signal that the down-move may be approaching its end.

2. (SC) Selling Climax: panic sets in and the stock displays widening spread as selling pressure causes high volume. This culmi- nates in climactic fashion. The final selling by the public is being absorbed by large professional interests.

3. (AR) Automatic Rally: the selling pressure is exhausted and buying in the stock easily moves prices up. Late comers to the short-side are also easily shaken and forced to cover fueling the move up.

(TR) Trading Range: prolonged range the stock will trade in while larger interests accumulate large positions towards the bottom half of the range. The public will become exhausted and bored with this process and eventually sell their interest prior to a major mark-up in the stock’s price.

4. (ST) Secondary Test(s): this is the base building process. Re- testing the bottom of the TR for supply where spread tightens towards the range where the SC took place.

(CREEK) Creek minor resistance that represents the floating supply in the upper edge of the Trading Range (TR). This supply must be absorbed before a meaningful move to the upside can ensue.

8 & 10. (Spring): occur later in the Trading Range (TR) and allows the large market operators to flush and shakeout the final supply which remains at the bottom. This testing process allows the larger operator to determine the stock’s readiness for mark-up. If there is little or no stock available at the lower end or under-cut of the Trading Range (TR) then the large operator knows the stock is ready. Moderate volume signals that continued testing of the range is necessary before significant mark-up can result. Heavy supply at this zone signals a further prolonged range or lower lows.

(JUMP) Jump across the creek: finally the stock has been prepared for mark-up and in a strong thrust the large operator buys his final line and begins to advertise his mark-up campaign by painting the tape in obvious volume.

11. (SOS) Sign of Strength: the advance is under-way. Volume expands and spread is wide. The floor brokers as well as speculators become interested in the shares and start to bid prices up.

12 & 14 (LPS) or (BAC) Last Point of Support or Back Up to the Edge of Creek: LPSs and BACs are the final stages of the testing process. Pullbacks (LPSs) occur on low volume as the operators square their position and speculators cash in on small stock fluc- tuations. Note that a series of SOSs and LPSs is good evidence that a bottom is completed and the mark-up has begun.

mlation and distribution is through a visual depiction. Wyckoff developed some 80-years ago.

I I -“‘5.I - iIzL(--o;J- I

-I-C+

DISTRIBUTION SCHEMATIC

1. (PSY) Preliminary Supply: is the point where heavy selling creates resistance and thwarts the up-move. Volume and spread widens and prices make little progress, this is the first sign that the up-move may be near finale.

2. (BC) Buying Climax: widening spread and a final thrust is accompanied by price failure signaling that the large operators have begun to throw shares of their line at the market. These orders are generally filled by the unsuspecting public interests at prices near the top.

3. (AR) Automatic Reaction: Public buying is quickly exhausted by this concentrated and intense selling as prices then fail miserably.

(TR) Trading Range: The range which the larger interests will work off the final shares of their position and begin to build a short interest in the stock.

4. (ST) Secondary Test(s): If a top is in place prices should fail at the top-end of the TR. As the larger interest work their line off, public demand pushes prices up briefly only to fail due to the lack of a concerted effort. Favorable news is used to lift prices as the larger operators liquidate their line.

10. (SOW) Sign of Weakness and (ICE) The stock trades extremely heavy as supply shows its dominance and pushes the stock to the lower end of the Trading Range ( TR). A line of demand exist at the bottom which forms the fragile ICE that holds the stock in the TR.

11 and 13. (LPSY) Last Point of Supply: The ICE gives the public a false sense of security as it is tested. Feeble rallies ensue and quickly fail. Demand is very weak as supply quickly depresses prices. This is the last chance for large operators to sell their final positions prior to the ICE giving-way.

11. (UTAD) Up-Thrust After Distribution: Although not as often, this is the sister to the Spring and Shakeout in Accumulation. UTAD’s are a final climactic rally which definitively tests the buying interest in the stock. They occur very late in the distribution phase and are often caused by weak short-sellers of a stock quickly looking for cover in what appears to be a resur- gence of the mark-up phase. A UTAD will rally in vacuum-like fashion through the top of the TR but, prices will quickly fail at the top and we will immediately fall back into the TR. The UTADs signals that the final distribution is over.

The ICE eventually cracks as the public “drowns” on supply. When rallies attempt to break above the ICE it seems this range is frozen back over. The public cannot break its head back through the ICE. Eventually, they sink to the bottom in dismay.


02/05/79

77 OEC ‘78 FEB NflR APR tlFlY JUN JUL AUG SEP OCT NOU DEC ‘79 FEB

- - - . - - - . . . . . . ,

77 FI M J J FI S 0 N 0 ‘78F M A V J J PI S 0 N 0 ‘79 F fl FI


FORD flOTOR CO. (1 x 1) 1 l/25/83

4 .--1 81 0 N il ‘62 : ri R N J J A S 0 N k il .i S 0 N

40

30

20


............... ........... .........

68 MTAJOURNAL/WINTER1994-SPRING1995

SEARS ROEBUCK CO (1 Y 1) 07/27/84

,_._................. . . . . . . .

81 0 N ‘82 PlAtlJJFtSON ‘83 PlfltlJJASON ‘84 flFIfl J J

..___......._____.___..........................................................................

79 SO N 0 ‘80 F tl fi fl J J A S 0 N 0 ‘81 F ll h fl J J A S 0


r


Wyckoff’s method of technical analysis utilizes a judgmental approach which allows the investor to annotate the chart and develop a frame work to operate within. By combining both models, more clues are available as to when it is prudent to assume the risk of an investment. In all, CANSLIM gives us clues as to where to locate the greatest companies and Wyckoff specifically tells us when to buy and sell them.

Hence, combining CANSLIM and Wyckoff as a total system includes many of the elements we find necessary in stock market speculation.

Combining Wyckoff and O’Neil results in:

1. Quantifying those characteristics most common in winning stocks and campaigns

2. Interpreting the law of supply and demand through regimented chart analysis of the trading range

3. Defining cause to pinpoint price objectives and proportionality through the use of Point & Figure charts

4. The interpretation of effort and result through Wyckoff’s descriptive chart annotations

CANSLIM surely identifies the effects that great new products, innovative new services, and new in- ventions have on a company’s performance. Wyckoff s discipline of position-taking, holding, and selling allow us to identify the potential stock price appreciation. This happens, early on, before a historic run is quickly made, and all of the superior characteristics of the company are discounted in the marketplace.

The study further leads us to conclude that major moves generally are preceded by significant accumulation zones and, often, periods of re-accumulation occur on the way up. Interestingly, this concept of a big base investing runs parallel with much of the analysis done by William O’Neil + Co. and, additional works published in the Market Technicians Association Journal. Two journal articles on big base investing that come to mind are those published by William S. Doane and Tom Dowse.

In sum, we believe that a system utilizing O’Neil’s searching capabilities and Wyckoff’s judgmental analysis affords the investor a complete methodology for growth-stock investing. We hope, after reading this work we are all better prepared to profit from the opportunities presented by the “Greatest Future Winners in the Stock Market.”

BIBLIOGRAPHY

Coppola, J.C., III, Harmonizing Wyckoff and CANSLIM, (un- published manuscript, 1994).

Comfort, M., “The Technical Securities Analyst Association,” William O’Neil + Co. Inc., Fall 1992 Presentation.

Doane, WS., “Broad Bottom Configurations and Their Application to Investment Strategy.” Market Technicians Association Journal, May 1984.

Dowse, T., “Big Base Investing for Long-Term Capital Gain.” Market Technicians Association Journal, February 1987.

Financial Information Inc., The Financial Stock Guide Services, (1993).

Flanagan, M.J., “Stock Market Timing An Empirical Evaluation.” Market Technicians Association Journal, February 1986.

Forte, J., “Anatomy of a Trading Range” Market Technicians Association Journal, Summer-Fall 1994.

Hutson, J.K., Weis, D.H., and Schroeder, CF., “Charting the Stock Market: The Wyckoff Method”, Technical Analysis of Stocks and Commodities, Seattle, Washington, 1991.

Merrill Lynch & Co. Global Securities Research & Economics Group.

O’Neil, WJ., The Model Book of the Greatest Stock Market Winners, 1953-83, William O’Neil + Co., Los Angeles, CA, 1983.

O’Neil, WJ., How to Make Money in Stocks, New York, McGraw Hill, 1991.

“NYSE & NASDAQ/American Stock Exchange Daily Graphs”, Daily Graphs, 1992.

Pruden, H.O., ‘Directed Individual Study”, Golden Gate University, Fall 1993.

Pruden, H.O. and Fraser, B., “The Wyckoff Method Lectures”, Golden Gate University, Fall 1992, Spring 1993.

Rhodes, WE., Merrill Lynch & Co., Global Economic Research and Economics Group, 1993.

The Wyckoff/Stock Market Institute, Phoenix, Arizona (602) 942-5581.

Vincent, B., Telescan.

Wyckoff, R.D., Wall Street Ventures and Adventures Through Forty Years, Fraser Publishing, 1930, 1985.

Wyckoff, R.D., How I Trade and Invest in Stocks & Bonds, Fraser Publishing, 1924, 1983.

Wyckoff, R.D., Stock Market Technique Number One and Two, Fraser Publishing, 1933, 1984.

CHARTS PROVIDED by Equis International, MetaStock 4.0.

J.C. Coppola III is the President of J. Philip Fund Management, Inc. a registered investment advisor in San Francisco, California. Mr. Coppola’s firm is the general partner of a top ranked hedge fund, J. Philip Fund Partners, L.P, which employs the strategy represented in this paper professionally in managing securities portfolios.


NOTES


44 - 1994 winter

Documents