2 2 bát Trang, 262THO-Activision
“This isthe reverse 4 Long 4-Shon B Entry Note the symmety ofthe Spread Banda in the Aeme leg ofthe ATVI-THQI pur trade shown in Figwe 24 andis Spread Indicator Essentlly, this spread is @ mor image ofthe spread shown inthe previous esample, but the SB valueis the same (0.1025) [Ea 3 26 Examples: “Table2.6 TradeStation traegy PerfermanceReport-AcneP System ATVLS min
[rota Net Prost 89,114.00 Open position P/ S00
|ceoss Profit $22.35200 Gross Loss (81327800
[Teal # oftrades 32, Percent profitable 93%
[Number winning trades 19 Number losing trades B JLargest winning trade $3,159.00 Largest losing trade (§2880 00)
JAveragewinningtmade S1,I7853 Average losing trade 102133) Ratio ay wing less HS Aygtrade (wm Sos) sosist
[stax conse, Winners 4 Mixcconses losers 3
Ave bars in winners 50° Avg bars in losers a
Intex intraday drawdown (85,149.00)
Profit Factor 169_ Mac contracts held 3.600
“Table227 TradeStation Strategy erformanceRepert- AcmeD SystemTHQES min
Figure 25 THQlncorporaed- Activision Pair
“ables 2.6 and 2.7 show the resus oftrading both es ofthe ATVI-THQL pi from November 6®, 2000 though March 22nd, 2002 The avenge trade netted 3285 forthe ATVIieg and $462 forthe THO le, totaling 8747 forthe stock pair Net of commissions and slippage, the average trade for this account Sze ‘Would fll inthe range of $500 to $600 This example atypical scenario fer pais trade: one profitable lee and ene unpositee lee The oa sto eaitalize onthe price diffrence with proper postion zing Some ofthe mest effective pir trades occur when one stock the pair gaps Aispoporionately tothe ether tock Fer exanle, ftw stocks with the same ‘ATR but diferent prices gap up ene point, eg the stock futures ae up, then the gap creates a pricing disp tne sock closed 30 and the ether at 10,
then the pair apread is G1 / 30) 1.033 minus @1 / 40) 1.025 equals +0.008
Hence, the former sock sovervluedrelativeto thelater,setingup apostle Short A Lone B pai ra
[real Net Prosi 381479700 Open position Pi $000
lores Profit $27,628.00 Gross Loss (1283100
[Towa # oftrdes 32 Percentprofitable 6050
‘umberwinningtrades 30 Number losing trades “
[Largest winning trade $3.360.00 Largest losing trade (8150060) laveragewinuingtrade — $I3SI4O Avengeleigtmde —- 106939 lRaiozegwinsvg les 129 Aygtrade Gvin & lass) 46241
conse Winners 3 Max conse: loses 3
Lave bars in winners 45 Avi bars in losers a INax intraday drawdown ($3,971.00)
[profit Pater 215 Mix # contracts held 2.600
Trang 2
st 2Pair Trading 263 Apache-Anadarko
‘The New York Sock Exchange is an excellent vehicle for trading correlated pals, especially among the cil service companies The APA-APC pair usually thas & correlation above 09 and exhibits smecth multi-day swings between the spread bands {In Figure 26, APA becomes relatvaly expensive to APC a the spread icks ‘the upper spread band This is an example ofa Short AcLong B Entry As soon asthe Spread creases back below the upper SB, we short 1400 shares oFAPA sĩ fa price of $758 at 1-00 pm and buy HO shares of APC at $800 Two hous later at 300 pm, the spread crosses below zero, the APA short is covered al 5748 fora profit of $700, and the APC long is sold st 58.10 for aprofit of SMO Inthis case, the total apread profit is $840 with one leg ofthe pair making: meet ‘ofthe profi De skeptical ofexceediagly narsow Spread Bande for sock piss with High ‘conations This is probably an indication that the correlation period is too small, Use 30 days at @ minimum-we recommend st least 90 days to fator in te quarterly earnings eyele The problem with narrow Spread Bands is that they are sensitive to small changes in He stock price, and too many signals wil ‘be generated Moreover, ifthe stocks do not fave high ATRs then overcoming the slippage and commissions will be difficult
we 26 Apel Avante Pan
26 Examples %
264 Allstate- Progressive
‘An advanced pair wading srategy uses in a specific stack sector One ofthe rock is growth a coclated poi ofa leader and alageard tock and the other tock is value stock or blue chip sock with ile woaility Typical, the bellwether stock sats moving oppesturity Here, two insurance companies have been chosen, eamervative before the ether, and the agin pice ation erates a spread Allate(ALL 2NYSE)andmomentum InFiguee 27, ALL stats declining at 1145 am and by 100 pm has become favorite Progsessive(PGR:NY SE) ‘undervalued waive to PGR When the spread erases above the ower SB, a Jong ALL signal is generated, and thus PGR isa short the spread is covered at the end ofthe day For the perio 11/06/2000 through 03/01/2002, each leg of
this spread had profit factors of2.32 and 321, rspoctivaly With 90 many trad- ers on theinformtion highway, it pays togo offroad and play niche strategy
‘When seanaing for stock pairs, Jock for the behavior demonstrated in the chart below While Allstate was dectiniag Progressive managed to stay up for several hours before filling, Essentially, we are looking for phase shift between Imo stocks so we recommend Keeping a quote table fall pairs to quickly judge ‘which pairs are showing price discrepancies
Oa pa ye oe om ee ee De ew HH Tes ee
gure 27 AI
Trang 3
56 2 Pair Trading 265 EmuleOloge
Tieme 28 is anillmaration of how a news itm cess opportunity On Jammary 23rd, 2002, QLogic Corporation reported earings afer the cling tell, nt then commented en "ited visbiliy" We were tacking the Enlex-QLagic ris, sothe flowing day Emulex stayedwell shove the upper SB allay tong At 5550 pm, aspke in OLGC exried the spread below the SB, where 70 shares of EMI were shorted Simultaneously, slong postion was taken in QLogic Just two bars ater onthe flowing monning, the spread depped below ze athe spreadwas covered ‘trader might question why QLogic would rise so suddenly The answers that picing dspaiies are creed near and atthe closing bel, aswell a tthe ‘opening bel Because ofthe anifciaity ofthe moves, these times provide the tes quick pair trading eppestnitis, Do net think aboot why he trade hap- pened just ke it because fie doesnot apy here The gical trader probably
got caught leaning the wrong way, major market index such asthe Nasdaq 100 cr SEP 500 With the growth in Some ofthe mest interesing pais are composed of socks thar belong to a
index funds and exchange-traded fimds (ETFs) money managers are obligated to
ty or sal he component stocks that compose the inex Inthe absence ofany ness, a index sock cannet geo far out ofwhackbeore index baying or sell- ing brings the stockback inline
26 Examples 37
266 RF Micro Devices-TiGuint Semiconductor
Trang 458 2 Pair Trading 2.7Pair Trading Strategies 59
2.7 Pair Trading Strategies
We keep an Excel spreadsheet of at least fifty correlated pairs for the Acme P System Of course, the stock market has an almost infinite number of possible pairs, but trading this strategy requires both liquidity and volatility, making most of them infeasible Filter out those stocks with little volume and low ATRs The key is to find two competing stocks in the same industry satisfying these criteria
- RF Micro Devices (RFMD:Nasdaq)-TriQuint Semiconductor (TQONT: Nasdaq)
- UnitedHealth Group (UNH: NYSE)-WellPoint Health (WLP:NYSE)
2./.1 Tips and Techniques
; 1 When legging into a pair trade, initiate the short trade first to make Following are some examples: sure you can get an uptick, unless you have a forward conversion or
- Abgenix (ABGX:Nasdaq)-Medarex (MEDX:Nasdaq) bullet that allows you to short without an uptick
- Arch Coal (ACI:NYSE)-Massey Energy (MEE:NYSE) 2 There are several alternatives for closing out a pair trade Some traders - Allstate (ALL:NYSE)-Progressive (PGR:NY SE) prefer to close out all trades at the end of the trading day Others pre- - Apache (APA:NYSE)-Anadarko Petroleum (APC:NYSE) rer to toni fee go to the opposite spread band over a period of ~ Activision (ATVI-Nasdaq)-THQ Incorporated (THQT:Nasdaq) 3 Ifone ofthe stocks in the pair trade has significant news, then the
- Altera (ALTR:Nasdaq)-Xilinx (XLNX: Nasdaq) pair trade may not be feasible because the spread may widen dramati-
- BJ's Wholesale Club (BJ:NYSE)-Costco (COST:NY SE) cally; however, the system rules preclude entering a trade until the
- Ballard Power (BLDP:Nasdaq)-FuelCell Energy (FCEL :Nasdaq) spread starts reverting to the mean
- Citigroup (C:NYSE)-J.P Morgan Chase (JPM:NYSE) 4 Find stock pairs that fit the "leader-laggard" criterion within a specific
Caterpillar (CAT:NYSE)-IngersoU-Rand (IR:NYSE) industry niche Do not settle on a general sector; look for stocks with products that directly compete with cach other, ¢.g., two companies - Carnival (CCL:NYSE)-Royal Caribbean (RCLNYSE) with a similar cancer drug
- Capital One (COF:NYSE)-Household International (HI:N YSE) 5 For intraday pair trading, use as small a time frame as possible The - Calpine (CPN:NYSE)-Dynegy (DYN:NYSE) window of opportunity can be small, so three-minute and five-minute
- Centex (CTX:NYSE)-Lennar (LEN:NY SE) charts should be used Another option is to change the Acme Spread
- Quest Diagnostics (DGX:NYSE)-Laboratory Corporation (LH:NYSE) indicator to alert the trader whenever either of the SBs is touched
6
DuPont Photomasks (DPMI:Nasdaq)-Photronics (PLAB:Nasdaq) EMC (EMC:NYSE)-Network Appliance (NTAP: Nasdaq) Emulex (EMLX:Nasdaq)-QLogic (QLGC:Nasdaq)
olamarn ach IN Woraan
5 Stanle MAY DN Home Depot (HD: NYSE)-Lowe's Companies (LOW: NYSE)
Human Genome Sciences (HGSI:Nasdaq)-Millennium Pharmaceuticals
(MLNM:Nasdaq)
Starwood Hotels (HOT :NYSE)-Marriott (MAR:NY SE)
Internet Security Systems (SSX:Nasdaq)- VeriSign (VRSN:Nasdaq)
Jabil Circuit (SBLNYSE)-Sanmina (SANM:Nasdaq) Jones Apparel (JNY:NYSE)-Liz/ Clairborne (LIZ:NYSE)
Mercury Intractve (MERQ:Nasdaq) Micromuse (MUSE:Nasdaq)
PeopleSoft(PSFT :Nasdaq)SiebelSystems(SEBL: Nasdaq)
When assessing a potential stock pair, research its technical character- flat-lining Look for spikes in the form ofthe letter "V" In contrast, some stock pairs are so correlated that they are not worth trading
News is not necessarily bad for a pair trade-it can create opportunity Unless the news is significant, the pair trade is a good option; other- wise, the reverse spread may be a better trade Psychologically, this is a difficult type oftrade but gets easier with experience
We prefer morning trades around 10:30 am and afternoon trades around 2:30 pm because these times tend to be intraday reversal points in the market Ifthe spread is still widening after 3 pm, then traders are probably caught the wrong way, and the spread will proba-
Trang 560 2Parr Trading
9 Compare the performance ofthe system using today's open versus yesterday's close Some stocks exhibit a tendency towards unusual gaps, and this usually improves pair trading performance
10 To adapt the Acme P system to position pair trading, add a parameter
3 Pattern Trading
that specifies the number of days to carry the position Then, multiply this number by the Spread Band value to widen the bands Finally,
change the Pricel and Price2 parameters to reference the closing prices of N days ago
11 We monitor all pairs with a set of TradeStation workspaces, six chart pairs per workspace Note that running multiple workspaces with multiple charts requires extensive computing power; we recommend a dual processor machine
Paranoia is the belief in a hidden order
behind the visible
Anonymous
Computer technology has changed the Zeitgeist of the market into a real-time information machine, sending multiple data sources flooding into a trader's computer Currently, raw data feeds supply primitive quote information The next generation of data feeds will feed patterns into the computer, mining data
on the fly to identify technical patterns that match the trader's criteria
A pattern combination system identifies charts with two or more patterns with a similar bias-bullish or bearish Ifa chart shows both bullish and bearish qualities, then the bias is neutral Essentially, the system catalogs all known bar patterns and then identifies combinations of them Furthermore, the system is extensible A trader can add as many patterns as possible, building the pattern catalog For example, Dunnigan developed a swing-trading catalog of patterns that could be implemented in a pattern combination system [9]
The system is not limited to any particular style of chart Every candlestick pattern has an analog that can be encoded in a normal bar chart [22, 23] So, a Harami can be combined with a Gann pullback [18] The exciting aspect of the Acme M system presented here is that the trader is free to invent his or her own patterns and plug them in Design the pattern, code the pattern function, and call it from the Market Patterns model As an illustration, we designed a sample bar pattern called a Cobra because ofits resemblance to a striking snake, with a head at the top and a tail at the bottom
Non-technical information about a stock can also be encoded in a pattern For example, a stock may have a quarterly earnings pattern cycle [13]; it may
simply be the datefive days before the company's earnings report [his earnings
date may coincide with a pullback pattern, so fundamental information can be
combined with technical information to ITCHto a trailing signal
Trang 6° 3 Pattern Trading,
‘oat or its EPS growth We envision the day when Internet-wide public feeds transmit all Kinds of nfomuation in real-time: earnings reports, baseball scores, ‘weather forecasts, đc,
3.1 Market Patiems
“The Market Patterns indicator iss series ofletters that identify specific bar pat- terms, Gautley inspired this idea because he made Liberal use ofthe letters ofthe alphabet to identify market pivots [15] For our purposes, a single letter denotes cach pattern, Forbush patier, aletteris placed on top ofthe bar For bearish paiterns, the leter is placed below the bar The list of Aeme patterns is shown, in Table 31 “Table3.1 Acrne Market Peters
Test
VZene ¬
31.1 Cobra(c)
‘A Cobrais a narrow range bar named for its resemblance to the snake, ie, the ‘open and close finish at eppeeite extremes ofthe bar For a bullish Cobra, the ‘open must be in the bottom oftie range, and the close must be inthe top ofthe rage For a bearish Cobra, the open niust be in the top ofthe range, and the ‘lose must bein the bettom ofthe range Figure 3.1 shows two bullish Cobras with the letter shove each bar Typi- cally, the Cobra precedes a large-range day The pattem is essentially 2 narrow range dey showing strength The letter N denotes a narrow range day; itis not ‘a independent pattern pec se but is noted when other pattems are present (see Section 3.2 on Pattom (uites 31 Market Pattems 6 igure 3.1 Cobra 31.2 Hook (H)
‘A Hooks a retracement bar that appears to reverse the retracement trend but ten elses in the direction of the retracement Figure 32 is an example of a bearish Hook with the letter H below the bar: a sesies ofpalback days, then the stock paps up al the open but cleses down on the day, contimaing the pullback
[Beventaly, the buyers here were “hooked” on the day's open
Trang 7
“ 3 Patter Trading
38.13 Inside Day?)
‘An Inside Day 2 bars an inside dy! followed by anetherinside day By defini tion, iis @ narow range day without any directional bia, thus the reason for ‘annotating both the top and bottom ofthe bar with the fetter IL A trade would normally be taken in the ditection ofthe breakout, asin Fignre 33 below Note that an inside day isnot necessarily a naurow range day Ian ouiside ‘ery large range, then the inside days range may sill be preaterthan the ATR day has a Figure 3.3 Inside Day 2 31⁄4TaI4L)
{A Talis a bar where the open and close re In the same half-bar Ifboth the ‘lose and open are in the upper halfof the range, then that is a bollish tai If ‘both the close and open are inthe lower halfofthe range, then that is @ bearish ta Figure 34 shows two bollishtails with the fetter L above each bes; each tail is also @ narrow range bar
“The Acme Range Percent function calculates where a price falls as apercentage ‘ofthe range (between 0 and 100) for a single bar or for'a number ef bars, this ‘umber s the Range Percentage (RP) Fora single bar, the low ofthe bar has an RP of 6, and the high ofthe bar has an RP of 100 For a range of bars, the lowest low ofthe range has an RP of0, and the highest high ofthe range has ‘an RP of 00
31 Market Paterns 6
‘The RP is also used as an input parameter to specify where the open and close ‘must appear on a given bar For example, an RP of 80 means that the open ‘and close must be in the upper or lower 20% ofthe bar The trader can adjust the RP for certain paterns, eg, $0 for more restrictive or 60 for les restrictive Ifthe trader wants shaven heads [22] then the RP can be set even higher Ui- rately, the choice isa tradeoffbetween discretion and automation tl TEE Figuee3.4 Tail
“The FasyLangnage code for identifying a tail is shown below in Example 31, Each patter function assigns the fllowing codes to return the pattern satus:
Trang 8“6 Pattern Trading enetail = 0;
RP'= (AcsetangePercent (Open, 1) + AeneRangePercent(Close, 1)) / 25 ‘Condationt = RP >= (1 ~ Peccent);
Condition? = AeneRargePercent(CLose, 3) »= (3 - RP25); IF (Conditions or Covdition2) and Benangepecen( Low, Raxgeargt) <- Percent nd Tae, đa SD rd 13) > P50 Then
"acnetail = 35
Conditions = RP <= Percent;
Conditions = AcnefangePerceat (Close, 1) <= P25; IF (Condit fonn or Cosditien2) and
AenetangePercent(High, RangeLength) >= (4 - Percent) and BeneRangePercent(Open, 3) € RPSD and
AenetorgePercent (Chose, 3) < RP50 Then ‘aewetail = 25 3⁄45 Baram 6
‘A Hanamiis clasäc candleglixk paVern Its atwo-bar pattern where an inside bar follows a wide range bar, and the inside bar closes in the opposite direction ‘A Haram by itself does not indicate a change in rend, but it can signal a rever- ‘sal when combined Vearish Harami ((he close is greater than the open in the wide rage bar fol- with other market pattems Figure 35 is a chast with both a Towed ty the close balow the open in the inside bar) and a bullish Haram (he ‘lose is less than the epen in the wide range bar, and the cose is abawe the open inthe second inside bas):
Tgguec 68a
31 MarketPatterns o
“The function Aemeffaram! implements an extended Hiarami pattem for up to ‘ree bars, that is, ifthe wide range bar is flowed by a series ofinside bars, then ‘he pattern isin effect until pice breaks out ofthe original wide range bat
346 Pullback(P)
‘A Pallbackis the traditional Gann pullback pattern, A bullish pullback eceuss ‘when the trend is up, and price makes a certain mumber oflower lows ora com- bination oflower lows anc inside days [4] A bearish pullback cocurs when the trend is down, and price makes a certain number of igher highs or a combina- tien ofhigher highs and inside days Figure 26 shows an esample of abeasish Palback withthe letter P below the bar
“The function Acme Pullback defines a paliback in cther ways, too Either a bullish Tail or a bullish Cobra ean mark the bottom of a retracement in an up ‘ond Similasty, a bearish Tail or Cobra can mask the top ofa retracement in a <downtrend The EasyLanguage code for the AcmePullback function is shown in Example 32: Example 3.2 Function AomePalback (9990142549144 2290148100841.040.4014400048.410.00414.40100 c6 ÄœreFullback: Search for a Gann Pullback Pattern ceethtsvasesennstecotenseasecsartavesseananrecoestoneeeonsqtenacss) inte meric), terethihimerie)s Variables: PullbackBars(3), RangeLength(s}, RangeFactor( 0.7); ‘AceePullback = 03 Conditions AcectetraceDow Pullbackars) = Tre
TF Aox(Length) >= AonLandt and
Daplas(Length) > Dalhirus(Length) ard (Gendittont or Confit fon? or Condit int) Then cent (Close, Rargclength) <- Porcent and hk
Trang 9
“ 3PattemfTrading
Pa — ——
1Í A0M(Lergtf) >= ADMLinit and
tRuhinas(LEngth) > OMPLus(Lengh) ond
AanofargePereert Close, Rangelength) >~ (2 - Percent) and (tem or Condition Condit) Then fenePullback = 2;
‘TheAcmePullback function is divided into two code ftagments: the frst section Tooks fora bullish pullback and the second section looks for a bearish pullback ‘A bullish pullback must satisfy the following minimum criteria:
1, The ADX over Length must be greater than or equal to the ADN init 2 The +DMI must be greater than -DMIL indicating an wp tren 3 The Range Percentage (RP) ofthe close must be 35% or less ver the
past Range Leng bars
4 ‘The bar pattern must be a bullish three-bar Pullbaek, Tal, or Cobra, ‘The minimum criteria for a bearish palitack are reversed for items 2, 3, and 4 ‘The -DMI must be greater than +DMI, and the RP must be at leat 65%
Figure 3.6 Pallback 317 Tee(1)
A Test 3 ha that breaks previous sappont on resistance and then rebounds in te other die Uist Vestal it Bae Beaks cow a prev he
(art Ie the ae Jes aed a hoes le ts hai ol
31 Matket Patems: °
‘eatsh Tes, the Test bar ewoods a previous high (cannot be the high one bar go) and closes inthe bottom halfofits range In Figure 37 below, the highest high isa bearish’Text masked bythe letter It breaks the previous high offour ‘ars age bat closes near the betiom fits range
Trang 10
m0 4 Patter Trading
“The 1°Zone is another Acme-desgnod pattern It is pattern based on a break cfweekiy support ccresstance, The V Zone is named forits resemblance tothe Jette *V* on the chart fora bullish pattern and an inverted "V" for beasih pal- tem (See Figure 38) The V Zone is the only patern that takes the weekend into effect because a stock that ches on its fw forthe week wil tod to break that low the following week Over the weekend, peopl in such a long pation {ink about how poody th position is faring and have two days to build up «motional capital in thee potion These emotions tigeer panic the fllowine week ifthe stock continues its dovenward mementum The BasyLanguage code for calculating the V Low price is shown in Example 33:
Example3.3 Function AomeV Low (9v trersrreorrrran — #kacMr Du the V bow Price .<ƯỮ ửừờ Ieputs: VolstilityFactor(tneric), Lêngth(Neneric);
AeneMUex = 0.0; Tf Loai(a) > 0 ang
Volatility (Length) > 0 Then ‘eneVton = Louk(a) ~ (Volatilityfactor * Velati1ity(Lergth));
‘The fornmla for calculating the V Low Zone is simple: determine the low ofthe previous week using the Lowi EasyLanguage function, and then subtract
muliple (olathsFacter) ofthe ATR from the weekly low Simitarly, forthe V High Zone, catentate the high ofthe previeus week with the ZZgfŸ function, and add the muiple ofthe ATR
32 Pattern Qualifiers
‘A Patiern Qualifier (PQ) is a characteristic ofa bar, not a bar paltern per se, however, it qualifies for the pattern count in the Acme M System For example, iftwo pattems are found, and the bar is an NR bar, then the pattern court is ‘equal to thee
32.1 NarowRange(N) “The first qualifies
the dat eis he Nerrow Range (N) qusifier,denotedby the fetter N on St placed on the chart unless other bar patane accompany it, aud at as marked only wh the be cf the bar pattern, ice, star placed ox only
132 Pattem Qualifiers n
‘one end ofthe bar For example, in Figure 39, the N qualifier is uppended to a bullish Tail (L) and a Test (1), forming the "LTN" sequence,
Figure 39, Nacrow Range Qualier 3.22 Average (A)
Trang 1172 3 Pattern Trading
Although the Average PQ is 50 bars by default, the trader can implement PQs for other critical moving averages such as the 200-bar moving average The key to pattern trading is to recognize those patterns that are important to technical analysts to supplement the patterns that the trader discovers on his or her own For example, a Fibonacci Pattern Qualifier could be designed to mark a swing retracement from a pivot bar ofa given strength
3.3 Pattern Trading System (Acme M)
With the patterns and qualifiers defined, we now implement a pattern trading system known as the Acme M System It simply counts the number ofbullish or bearish patterns on any given bar Ifthe number of patterns is greater than the minimum number of patterns specified by the system, and the swing trading criteria are satisfied, then a signal is generated
The swing trading criteria are two-fold First, the bar must be a swing high or low over a certain range ofbars Second, the swing high or low must fall in a certain range percentage over that same range Currently, the default value for the reference range is five bars, and the default value for the Range Percentage is 40% Note that the Acme M System avoids gaps that are not in the direction of the trade For example, if the high of the current bar is less than the low of the previous bar (gap down), then a long trade will not be taken
3.3.1 Long Signal
3.3 Pattern Trading System (Acme M) 73 Exit Rules: Profit Target
1 Sell half ofthe position at or above the High + (ProfitFactor * ATRy4) 2 Sell half ofthe position at or above the High of ProfitBars ago +
(2 * ProfitFactor * AT) Exit Rules: Stop Loss
1 Sell all shares at or below the Lowest Low for StopBars - (ExitFactor * ATR,4)
3.3.2 Short Signal
Calculations
1 Total the number of Bearish Patterns and Pattern Qualifiers 2 Calculate the Highest High of the previous 4 bars (HH,)
3 Calculate the Range Percentage of the Low ofthe current bar relative to the range ofthe previous 4 bars Entry Rules Calculations 1 Total the number of Bullish Patterns and Pattern Qualifiers 1 2 3 RPiy4 >= 60% 4 Sell Short the next bar at or below the Low - (EntryFactor * ATRy)
Calculate the Lowest Low ofthe previous 4 bars (LL,)
3 Calculate the Range Percentage ofthe High ofthe current bar relative to the range ofthe previous 4 bars (RPy,)
4 Calculate the ATR for the last 14 bars (ATR,,) Entry Rules
1 Number of Bullish Patterns >= 3
2 Low < LL,
3 RPyy <= 40%
4 Buy the next bar at or above the Hligh + (Entry Factor * ATRy,)
Exit Rules: Profit Target
1 Cover half ofthe position at or below the Low - (ProfitFactor * ATR 4,4)
2 Cover half ofthe position at or below the Low of ProfitBars ago
(2 * ProfitFactor * ATRy4)
Exit Rules: Stop Loss
Trang 12” 3Patem Trading ‘The Acme M code is divided into five general sections
= Trade Filtering
= Market Pattern Identification = Pattern Qualifier Identification ~ Position Sizing
= Signal Generation
‘The Acme M system uses two trade filers: Minimum Price and Minimum ATR The filters canbe applied individually rtogether For example, fter cut all socks with a price less than $20 and an ATR offess than one To wse just ‘ne filter, set the other filter to zero Fee example, st the Minimum Paice to ero, and st the ATR to a mumber Ifthe stock passes the filtering proces, then the eight market patterns are such as 15 tested in sequence Two strings are created that track the tilsh and bearish patterns foreach bar ifthe Acne pattem funtion returns 1, then tie patterns bulish, andthe cocesponding lets is appended to the ZoSing vale If the Acme function returns 2, then the pattem is bearish, and the letter is added tothe ShortSirng vase ‘Alter pattern identification, the baris tested for pattem qualifiers Fis, the ‘AcmeCnsverage function is elle to determine whether or net the bars on the ‘0-day moving average Ifso, then the letter A is added to beth the LongSming ‘and ShortString Then, all ofthe marrow range patterns are tested, any mare range pattern is found then the eter N is appended to the appropiate súng, depending on wher or nt a bar pattern has aleady been found For example, ife bullish Test pattern has been found, and te bar isan NR bar, hen the leer sequence “TX will be contained in the LongString vaviable Ifa bears Test aller has been found, then the fetter sequence "TN" wil be contained in the ‘ShortString variable ‘Now, with the pattems and qualifiers identified, the system compares the number ofpattems tn each ting withthe minimum number specified as an.in- put parameter, MininumPesers the LengString has a length greater than {he minimum, den apetntial lng entry has been identified ihe Shest Sting hha alengthgreaterthan the minimum, then a shest entry has been identified Filly, the system test the fllowing entry conditions before validating the sigoal = Swing condition,
+ Range percentage condition, and = Gao condition
“The EasyLanguage code for the Acme M System is shown in Example 34,
33 Pattem Trading System (Acme MD sample 3.4, Acme M Syston,
‘ese H Systen: Look for pultiple cotbinations of market patterns tem a ae 5 Harani (M) Pestiaa 7 Test (1) aE, qualifiers 1 Noving Average (A) 21 Narrow Range (0) Its: i Parancters) TRndeBsttes(3) {iter Faroneters Fittersontce)e iglesia, RhnieebrieeDs); Rhrie-AHC 0), (Gosttion Farsnters) Rao),
Trang 13% 3 Pattern Tea grercent(0.25), Gangetacter(0"8), tang), Uercent(0-3), Perea (0:2), Length), eons), Pength(ia), r5), NHàctor(G.T, Mega) MenEtI2Q0), ME), ightext(o), LowText(0);
a fentryFactor ATR); 1 ATRLength) Sertsap "tow = eneyrcter + MÔ;
‘hun trade filters)
TẾ Eilterson Then ‘TadeFilter = Close > HinaurPrice and ATR >= HinisurATE; Ế Trađefilter Then Eegin
Patternstring = “5 Tạ Ung ‘Shortstring = “5
ReturWalue = AeneCobra(CPercent, CRangeFactor); 1 Returalue > 0 Then Begin ĐạttemStrirg = "C's Tf RetucrWalve = 1 Then
LongString = LongString + Patternstring Else ‘Shortstring = ShortString + PatterrString;
End;
ReturWalue = Aemeook(HLength, Wrercent); li TH ter Ben Pattemstring =
TW hetunalie » 1 hen Longstrirg = Longstring + PatterString Else Shortstring ~ Shortstring + Pattoenstring:
313 Pahem Trading System (Acme M) Longstring = LongString + FatternStx ShortString = Shortstring + Patterestring; ReturnWalue = AceeTail(LPercen:, Length); Tf ReturnWalve » 0 Then Begin
PatternString = "0 TF Returralie « 1 Then
gạt ~ Legstring + Patterestring ‘Shortstring = ShortString + Pattemnstring; ReturnWalue = Acneisrani Tf Returnvalse > 0 Then Begin
Patternstring = "W's
Tf ReturtWalue = 4 Then LongStríng = LongString + Patternstring else Shortstring = ShortString + PatterrStrings End;
ReturmNalue = AcePullback(PADK, Plengt TẾ RetufrValue > 0 Then Begin Patternstring = “P's
nelensteeg ~ lơ tín « Ptterrstring ‘Shortstring = Shortstring + PatterrStrings Ens; Returnvalue = AeneTest; Tf ReturrWalue >'0 Then Begin PatterrString = "P2 Tf Returnalue ~ 1 Then dụ Dự Sung + Fatersting ‘Shortstring = Shortstring + PatterrString;
If High ›= AcmeVflgĐ(VEactor, FITterLength) Then ShortString = Shortstring + "V's 1 Low <= Reme\ton(VFactor, FilterLength) Then LongString ~ Longstring + "V's
(Pattern Qualifiers)
Trang 14Pattern Trading shortsteing = ShortString + PatferrString; tnd; Conditions» Acetarrangrtergth, 1) mở Ac®elorrodange(NIengtla, 0) Condition? = Low > Low{s] ard Binh < HigÖ[1] anổ Loni] > Lowf2} and HighLa] < KiệM[2]; Conditions = Aeneharrew@ange(MLength2, 0); {Conditions = AenelnsideDayii(Mergth3, 0)? Conditions = Range <= NRFactar * ARS
1 (Conditions or Condition? or Condition or Conditions or Conditions) Ther Begin Patterstrirg = "3,
TSR OTe cg longstairg = LongString ¢ Patternstring, 1 Strlen(ertStieg) 8 Thee
eng, DSI < MefSưing v attesrings
{calculate shares based on risk sodel)
Wo RenetetStares(Equity, Riskiodel, RiskPercent, RISKATR); {Wultiple Pattern Buy Signal)
TẾ Strlen(LongString) ›- HindmunPattems and Low < Lowest(Low, Mange - 31] and
‘enckangePercent(High, ange) <= MeangePercent and High > Lows] Then {raw Entry Targets on the chart)
Tf Dravtangets Then CConditiend = AeneEntryTargets("H", BuyStep, 0, 0, 0); fuy("*hene LE MP) N shares Next Bar of BuyStop stop; Em;
(Itiple Pattern Sell Signal)
1f strenStortstirg) >= Minimaattenns and nh saghest (high Marge = 13)
Khang leo, Meyc) v ( Mmgefered) mở Tốc Hạng] hen In ew etry Targets co the Gart)
1 Dramargts then Condit rd = Acnrtryorpeta(", 0, 0, Statstep, 6); seule Sk Hr} Sates Nest Har 6 Shortstop Stops
34 Bowles k
34 Examples
“The following chasts axe exenples oftade generled Ủy (he Acme M System Each example uses Equity of $100,000 and the Percent Volatility Mode! with = riskof2%e, For stocks, trae filtering is wmed on, and for melces, trade filtering is tumed off We do not apply trade filtering to the indices because they have lower ADX readings,
8441 Abgeni
‘The chart in Figure 311 shows an Acme M long entry on Angust 20°, 2001 “The number ofbuilish patterns is equal to four
~ A Hook (H) where a short would have been triggered Dut the bar reversed,
= A Tail (L) where the open and close are in the high end ofthe range, = A successful Test (T) of the previous Jow two bars eacter, and ~ The bar is a marrow range (N) bar
“The orginal position size ofthe entry was 800 shares Halfofthe position (400 shares) was closed two days later, and the rest ofthe positon was sold on the third and firth days ater entry
Trang 15© 3Patem Trading 342 PMCSiera
‘The chart in Figure 312 shows an Acme M short entry on January 25%, 2001, with the mmber oftearish patterns equal to three:
= A Hook (H) down after a three-bar pullback, and
= ATal (L where the open and close are near the low ofthe range, and ~ An unsuecessful Test (1) of te previous high three bars caer “The original postion size ofthe entry was 200 shares Beth halves ofthe posi-
tion (100 shares apiece) were covered on the following day for almost 40 pois “These abrupt reversals of fortune tend to occur before an earnings selease Bad ‘news just happens to leak out, awonderfally human aspect ofthe business,
Figure 3.12 PC-Siena Panera
[Note how the stock went in the CAN alter the iitial short entry was triggered ‘On the subsequent bar, the stock moved below its S0-day moving average, and ‘closed near the Jow of te day This chart exemplifies a confluence of bearish palterns and only increases the probability ofa winning trade The trader may choose to make the Acme system less restictive by removing the swing cxite- ria and just focusing on these pattern clusters, Aguin, the trader mskes the choice cfwhich stocks will be eliminated by the scans and how many need to be reviewed by eye As will be demonstrated in Chapter 9, the human eye and brain ‘are not always better judges of charts than the computer
34 Examples sỉ
343 Check Point Sofware
‘The chat in Figue 213 shows an Acme M long entry after a double hook Fins, some longs wete sucked in at atick above the high on the 13%, and then some shorts were entrapped a tick below the low onthe 14* (nt good idea on wide range tar) After the trade was triggered, the ATR factor prevented the trade from getting stopped eat atthe previous days tow on the 18 Curenly, the better methodology seems to be fading the swing traders around the lows snd highs, i there appear to be more hooks, proving thal just 18 apaticularsyl oftading becomes popular among the masses, the les ily it will workin the fuute Consequently, the ATR foctor on ents eiminates many ofthese types oftrades “The Acme software defines mechanical trade es forits systems; research hha shown tha two steratve systematic metheds work equally wel:
‘2 Exit the Jong position when the close is below the open, and exit a shot position when the close is greater than the open This techiique ‘works aftera swing ofseveral days because a trend reversal may be
oxthcoming beating traders who operate around highs and lows 18 Do net use profit targets after a holding pesiod of'our ot five days If
Trang 162 3PatemTrading 3.44 New York Futures Exchange
‘The NYFE chart in Figwe 3.14 shows an LE M signal Fist note how the stock has bounced off double bettom and is forming triple tp Second, the hat has formed a big "W", similar to the crisscrossed palm tres in I's @ Maa, ‘Mad Mad Mad World’ This balish pater portended @ major market rally in
‘erly March 2002,
Figure 3.14, NYFE Index astern
Jn Chapter 5, we discuss geomet formations suchas triangles, double bet- toms and tipse tops The purpese of covering these pattems is to comwey as rnuch information as posible cn the char In the example above, the double oto fom ealy to fate Febmuary provided supporting evidence forthe long trade entry Forthe trade exit, the approach to ariple top was ache that the up {rend may sill ave been infect, and the position could have been held longer tose ifthe index could breakout above the triple top ‘Strdsman uses the term “indicator pling” to espress the danger of wing Similar indicators tha give the trader a false impression ofa sure winner [30 Further, Bollinger tresses the need frindependent variables such as volume confi trades [1] When designing or implementing pater catalog, sive fora diversity of patter, each with « unique conc Atm dua ak, bata War ‘at Reva 34 Examples = AS Comverse Technology ‘The chart shown in Figure 315 exemplifies the "CAN do* spirit: 2 Cobra above ‘the moving average precedes a 156 move in four days Figure3.15 ConverseTechnologyPatiem 346 Nasdaq Composite Index
“The chat in Figure 3.16 is notjust for accountants,
Trang 17
a 3 Pattern Trading, 347 Computer Associates
Look: for stocks bucking the market trend, as the Computer Associates chartil- Iusrates in Figure 317 Before the gap down, rumors were swisling about the ‘company’s accounting practices during the Enron scandal Two days later after the trade entry, several major newspapers reported that the ecmpany was under investigation by the SEC
Figure3.17.ComputerAssociatesPattern
4 Float Trading
bcs mst work tongh the brains earlthe cams efgood coe bree men
cer they ave no betier he dreams RalpliWaldoEmerson
fanyone ever asks you why the Internet stocks soared as high as they did in the ate 19908, tell them it was the “low floats This was a group of stocks with geod earnings and revenue momentum, but more importanily, the demand out- stripped the supply ofshares Because the companies had ouly recently gone public, ew shares were available for trading (the flea; some companies issued uly three million shares, the equivalent of 300 investors owning, 10000 shares piece Compare this number of shares with established companies: such as General Electric, as shown in Table 41
‘Table4.1 Compasy Float Values
ssc somiet_| Float
cE 984 billion shares CIEN 301.0 milien shư
PDLI 355 millon shares THO 179 milion stares 8 million shares
As the prices peatedly forced to buy in theirshares, the companies declared of internet stocks soared, fucted by the short sellers who were re- stock spits These splits increased the number ofshare’, thereby increasing the float After two ot three splits, eventuslly these companies had floats of fify to a hundred million shutes With rices in the hundreds these companics attained market caps ‘eteater than the Iugest companies in Americ, in some cases ely cen bil