An early predictor for stocks?

4002

ce

d/v

on

.technicalanalyst.co.uk wwwThe publication for trading and investment professionals

The January BarometerThe January Barometer

Outlook for CAD Fine-tuning Fibonacci

Identifying when reversals are likely

Behavioural Finance

Investor psychology & the charts

Long-term viewremains bearish

Techniques

November/December 2004 THE TECHNICAL ANALYST 23

BackgroundThe pattern concept is that of a briefpause in market action as defined by adecrement in volatility, followed by afalse directional move. Volatility ismeasured simply as the (hourly, daily, orweekly, etc.) high to low range.Academics might say the hikkake playson the short-term mean reversionproperties of markets, by identifyingsituations where prices have beenstretched past their short-term equilib-rium value. Momentum is generated inthe early stages of the reversal path,causing prices to return to and eventu-ally through their prior equilibriumlevel. Technicians on the other hand,should recognize the hikkake pattern asa quantified compressed-time versionof the traditional "shake-out" pattern(see Figure 1).

The rationale for why the hikkakeconcept should have any effectivenessmay be due to the behavioral tenden-cies of small traders. Research by

Goetzmann and Massa suggest thatindividuals (i.e., the public) are predis-posed to "chasing" performance on ashort-term basis, and are equally predis-posed to reversing their decisions whenthe market moves against them. Inother words, individuals tend to beshort-term performance trend follow-ers. In the case of the hikkake pattern,perhaps it is the initial breakout thatattracts smaller participants into the

market. Another source of buying andselling comes from technicians and sys-tem traders. Buying and selling aroundpredefined narrow ranges is anextremely popular concept that goesback at least to the early 1900s (seeFigure 2). In fact, strategies based onthis idea gained in popularity followingarticles and books written by fundmanager Toby Crabel in the late 1980s.

However, if prices fail to gener-

QUANTIFYING MARKET DECEPTION WITH THE HIKKAKE PATTERN by Daniel L. Chesler

A savvy old timer once described Wall Street as "the only place

where they put prices up when they are having a sale." Indeed,

deception is a feature of most competitive fields including politics,

biological systems, financial markets and sports. Take for instance

the dummy move commonly used in football. The idea is to move

in one direction, thus unbalancing your opponent, before moving

away quickly in a different direction. The hikkake pattern is the mar-

ket's version of the dummy move. Hikkake is a Japanese verb

meaning to trap or to ensnare. In western terminology, the proper

name for this pattern would be inside day false breakout.

L O N G B A S E W I T H S H A K E - O U T M O V E

F a l s e B r e a k

Rally To

New Highs

Figure 1. This graphic, from William Jiller's classic 1962 book on charting, illustrates the tra-ditional style "shake-out" pattern, which has also been described by authors such asSchabacker, Wyckoff, and Edward & Magee. Hikkake patterns differ from this traditional"false move" schematic primarily in respect to time.

ate the expected trending behavior fol-lowing a breakout, a potential pool of"trapped" new commitments is created.Increased buying or selling pressurefrom the unwinding of these losingpositions is one possible reason whyprices continue moving in the oppositedirection of a failed breakout. Anotherpossible source of fuel propelling themarket after a false breakout is theopening of new positions in the direc-tion of the true trend. For example,many popular breakout strategies advo-cate placing stop and reverse ordersafter an entry.

Description The basic hikkake pattern is formed bythe combination of two price bars (twohourly bars, two daily bars, two weeklybars, etc.). The first bar must be an"inside" bar, which is defined as any barthat is completely encapsulated by theprevious bar's range. The second bar inthe pattern must have both a higherhigh and a higher low than the previous(inside) bar for a bearish hikkake set up,or a lower low and a lower high thanthe previous (inside) bar for a bullishhikkake set up. Note that the hikkake

pattern ignores the positions of theopen and close, known as the "realbody" in candlestick terminology. Thisis not atypical; other traditional candle-stick patterns such as tweezers, hang-ing-man lines and hammers also ignorethe open to close relationship.

Once a hikkake pattern has formed,the pattern is confirmed after pricesreturn up through the high of theinside bar (for a bullish set up) or downthrough the low of the inside bar (for abearish set up). Confirmation alsoserves as the entry trigger when usingthe pattern for trading purposes.Normally I look for confirmation tooccur within three bars following theinitial, two-bar hikkake pattern. Uponentering a position, risk is defined byusing the highest high (for shorts) orlowest low (for longs) within the pat-tern as a stop out point. This is not nec-essarily the low (or high) of the falsemove following the inside bar, since thelowest low (or highest high) can alsooccur anytime within the three bar peri-od allotted prior to entry. The testresults presented in this article arebased strictly on the foregoing entryand stop placement rules.

ExamplesUnlike most price patterns that fall intoeither "continuation" or "reversal" cat-egories, the standard hikkake patternplays both roles equally well dependingon where it occurs within a trend. Inorder to help the reader understandhow one pattern can play two very dif-ferent roles, bullish chart examples willbe presented. All charts are from thecurrent year. Bearish examples will notbe presented due to space, but are mir-ror opposites of the bullish examples.

In April, June bond futures were in apowerful downtrend. The market wasmaking new multi-month lows, long-term moving averages were pointingdown, and directional movement stud-ies were showing that the strength ofthe downtrend was actually increasing(these indicators are not displayed forsake of clarity). On April 26 an insideday formed, giving us the first half of apotential hikkake pattern (leftmostpoint 1 in Figure 3). The next day(point 2), a bar with both a higher highand a higher low formed, completingthe bearish hikkake setup.Confirmation came quickly, on April28, as the market traded below the lowof the inside day. A second bearishhikkake pattern formed between April30 and May 3, and was confirmed whenthe market fell below the low of theinside bar on May 4. Both of theseexamples were found in the context ofan existing downtrend and demonstratehow a bearish hikkake pattern func-tions in a continuation role.

Next we will look at an example of abearish hikkake functioning as a rever-sal pattern. The EUR/USD currencypair peaked in early 2004. After correct-ing down into May, the EUR/USD

Techniques

24 THE TECHNICAL ANALYST November/December 2004

Figure 2. This graphic illustrates a technique for trading based on the narrow-range phenome-non (volatility expansion-contraction). From A. W. Wetsel's (Wetsel Market Bureau, Inc.) ACourse in Trading, 1933, with permission from Donald Mack.

Techniques

November/December 2004 THE TECHNICAL ANALYST 25

began a gradual, low-momentumrecovery. On July 15 an inside barformed (point 1, Figure 4). The nextday, July 16, the market formed a high-er high and a higher low (point 2), ful-filling the requirements for a bearishhikkake. Confirmation came within twobars, on July 20, as the market fellbelow the low of the inside day.

Testing method and resultsTests were not aimed at determiningwhether the hikkake has any value as astand alone system, but rather to findout how often confirmed hikkake pat-terns represented valid false breakouts.The ensuing moves were counted andcategorized by their magnitude. Thisinformation should tell us whether thehikkake pattern adds value as an analyt-ical tool and can be used to guide ourentry and exit points.

For each valid hikkake pattern that isconfirmed within three bars followingpattern formation, a trade is enteredeither long or short. The amount ofcapital risked on each trade is definedas the difference between the highesthigh (or lowest low) of the confirmedpattern and the entry price (either thelow or the high of the inside bar).

The test uses a static protective stopthat remains in its original location anddoes not trail up or down as the tradematures. The test then counted howmany times the ensuing move equaled:

Figure 4.

Figure 3.

Less than 100% of initial riskbefore being stopped out.100% to 200% of initial riskbefore being stopped out.200% to 300% of initial riskbefore being stopped out.Greater than 300% of initial riskbefore being stopped out.

1.

2.

3.

4.

Techniques

26 THE TECHNICAL ANALYST November/December 2004

This information should give us someidea as to the average magnitude ofmoves following a confirmed patternand gives us some basis for settingprofit targets and for moving up stopson open positions.

Markets tested include the Dow JonesIndustrial Index, Russell 2000 Index,CBOT Bonds, COMEX Gold,NYMEX Crude and Natural Gas,USD/JPY and EUR/USD currencypairs, and individual equities GeneralElectric and Microsoft. Test results do

not include trading costs or slippage.TradeStation daily data was used for alltests. Continuous contracts were usedin the case of futures markets, whichmay or may not accurately reflect theresults of testing on individual futurescontract data.

The results of the test show that overfairly lengthy periods, ranging fromeight years in the currency pairs totwenty-four years in gold, the hikkakepattern produced an average win lossratio (number of winning trades versusstopped out trades) of slightly betterthan 1:1.

Table 1 shows the combined resultsof all trades long and short. However,we note from data compiled duringtesting, that winning versus losingtrades for some markets faired muchbetter than the average combined per-formance, such as long natural gas(3.2:1), short USD/JPY (3:1) and shortRussell 2000 (2.6:1). Less impressivemarkets included short Dow JonesIndustrials (0.9:1), short bonds (1:1),short EUR/USD (0.9:1) and longUSD/JPY (1:1). (A full breakdown of

individual long and short results isavailable from the author).

Research ideasDue to its simplicity, the hikkake pat-tern lends itself well to adaptations.One version, which I have found effec-tive at identifying trend reversals, takesthe standard pattern and applies thefollowing set of requirements to thebar immediately preceding the "inside"bar:

Note that since this version functionsprimarily as a trend reversal pattern, itoccurs less frequently in the data thanthe standard hikkake pattern.

The hikkake pattern is not limited totraditional bar or candle charts. Pointand figure and tick-based charts, whichuse a variable time axis, also exhibit

Figure 5.

RATIO

200+300

PERIOD MARKET STOPPED 100%-200% 200%-300% >300% WIN/LOSS VS 100

01/02/90 - 08/15/04 DJI 64 21 14 42 1.20 2.72

01/02/80 - 08/15/04 GOLD 131 52 32 88 1.31 2.33

07/01/94 - 08/15/04 RUSSELL 2K 22 15 7 26 2.18 2.35

01/01/90 - 08/15/04 BONDS 60 26 7 48 1.35 2.17

01/05/98 - 08/15/04 USDJPY 22 8 10 18 1.64 3.70

01/05/98 - 08/15/04 EURUSD 28 12 4 19 1.25 2.02

04/30/90 - 08/15/04 NAT GAS 37 27 10 38 2.03 1.85

08/14/84 - 08/15/04 CRUDE 68 32 18 45 1.40 2.01

01/02/90 - 08/15/04 GE 44 17 5 23 1.23 2.25

01/02/90 - 08/15/04 MSFT 41 13 13 32 1.41 3.57

LONG & SHORT COMBINED STATISTICS

Table 1.

The bar must close at the top ofits range (for top reversals) orthe low of its range (for bottomreversals).The bar's range must be lessthan the range of the previousbar.

1.

2.

Techniques

November/December 2004 THE TECHNICAL ANALYST 27

hikkake patterns. Figure 5 displays anexample of both a bearish and a bullishhikkake reversal pattern in theEUR/USD, from May and Septemberrespectively.

Final thoughtsI have a prejudice that says most tech-nical indicators and methods are simplydifferent ways of reflecting the same,or very similar, information. The com-mon threads that bind indicators, pat-terns and methods are therefore moreinteresting to me than are their appar-ent differences. In my own work, I havealways attempted to build models thatisolate these common elements.

My first attempt at this resulted in aless subjective way of identifying classi-cal chart patterns by distilling patternsinto separate volatility and cyclic com-ponents (see references). By compress-ing the traditional, detail-rich view ofclassical chart patterns down to justtwo main components, I discovered away to increase the method's robust-ness and improve its ability to general-ize for unseen cases.

What patterns and concepts were dis-

tilled in order to arrive at the hikkakepattern? Does the hikkake subsume theminutia of other approaches? Earlier inthis article it was shown how thehikkake concept overlaps with tradi-tional shakeout patterns. Elliotticiansshould find much in common betweenhikkake reversal patterns and fifth-waveterminations. Lastly, oscillator enthusi-asts will find that momentum diver-gences on a lower time frame chart willoften correspond with hikkake rever-sals on a higher time frame.

In sum, chart patterns put traders ina position to capture outsized moves,while giving a defined structure for set-ting risk parameters and exit targets.Patterns help us develop a case foreither a bearish or bullish outlook inconjunction with other inputs. Chartpatterns add context that statistics donot capture. The hikkake pattern servesall of these purposes.

Daniel L. Chesler, CMT, CTA, is acharter member of the AmericanAssociation of ProfessionalTechnical Analysts (AAPTA). Heprovides strategic technical fore-

casts for commodity and financialmarkets to proprietary traders, riskmanagers and brokers. E-mail:dan@charttricks.com

ReferencesToby Crabel (1990), Day Trading with Short TermPrice Patterns & Opening Range Breakout. TradersPress, Inc.

William L. Jiler (1962), How Charts Can Help Youin the Stock Market. Trendline, Division ofStandard & Poor's Corporation.

Daniel L. Chesler (1997), "Identifying SignificantChart Formations", Technical Analysis of Stocks &Commodities, Volume 15.

William N. Goetzmann and Massimo Massa (2003),"Index Funds and Stock Market Growth", Journalof Business, Vol. 76, No. 1, 1-28.Richard D. Wyckoff (1910), Studies in TapeReading. Fraser Publishing.

Richard W. Schabacker (1930), Stock Market Theoryand Practice. B. C. Forbes Publishing Co.

AcknowledgementThe author wishes to thank Yohey Arakawa,Associate Professor of Japanese, Tokyo Universityof Foreign Studies, for his help in finding the rightJapanese verb to describe this pattern.

UNLIKE MOST PRICE PATTERNSTHAT FALL INTO EITHER CONTINU-

ATION OR REVERSAL CATEGORIES, THE STANDARD

HIKKAKE PATTERN PLAYS BOTHROLES EQUALLY WELL .