Bollinger Bands Put To The Test PDF

Preview

Summary

uso de bollinger...

Description

BollingerBandsTradingStrategiesPutToTheTest

 ThisarticlelooksatfourBollingerBandstradingstrategiesandtestssomebasic ideasusinghistoricalstockdata.Itwasoriginallypublishedatjbmarwood.com butisreproducedherewiththesystemcodeincludedattheend. BollingerBandsareausefulandwellknowntechnicalindicator,inventedbyJohn Bollingerbackinthe1980s.Theyconsistofasimplemovingaverage(usuallythe20 period)andtwoupperandbottombandswhichareplacedanumberofstandard deviationsaway(usuallytwo). Theyarethusabletocapture9095%ofasecurity’spricemovementandtheyarethe perfectvehicleformeasuringvolatility. Whenthetwobandsarefarapart,thestockorsecurityishighlyvolatile,andwhenthe bandsareclosetogethervolatilityislower.BollingerBandsarethusthebasisformany differenttradingstrategiessuchastheBollingerBandssqueeze,theBollingerBands breakout,BollingerBandsreversalandridingtheBollingerBandstrend. ThenextimageshowstheBollingerBandsoverlaidonapricechartwithgreenandred arrows.Thisisatradeexampletakenfromstrategynumbertwo,whichyoucanread aboutbelow.

jbmarwood.com

 BollingerBandsParameters AsImentionedabove,thedefaultistousethe20periodsimplemovingaverage. Theupperbandisthenplaced2standarddeviationsabovethe20MAandthelower bandisplaced2standarddeviationsbelow. Ofcourse,youareallowedtouseanyparametersyoulike,butyoushouldremember thatifyouchangethenumberofstandarddeviationsto2.1,themovingaverageperiod shouldchangeto50.Likewise,ifyouchangethenumberofstandarddeviationsto1.9, themovingaverageperiodshouldchangeto10. BollingerBandscanbeusedonanytimeframeandareusefulforpatternrecognitionor combiningwithothertechnicalindicators. IlikeBollingerBandsalot,becausetheyareexcellentformeasuringvolatilityandthey canbeoverlaidonanytimeseriesnotjustastockprice. BollingerBandscanbeusedonvolume,openinterest,sentimentdata,almostanything. Theyarethereforeveryflexibleandallowtradersaquickwaytousestandard deviationsintheirmodels. BollingerBandsTradingStrategyOne AsIsaidbefore,BollingerBandsareanexcellentindicatorbutonlyifyouusethem correctly,andtheinventor,JohnBollinger,createdanumberofrulestoguidetradersas tohowtousethem.Youcanseethefull22B  ollingerBandsruleshere.

jbmarwood.com

OneoftherulescreatedbyJohnBollingeris: 8.ClosesoutsidetheBollingerBandsareinitiallycontinuationsignals,notreversal signals.(Thishasbeenthebasisformanysuccessfulvolatilitybreakoutsystems.) Inotherwords,ifasecurityclosesabovetheupperbandyoushouldseethisasa breakoutandgolong.Conversely,ifasecurityclosesbelowthelowerbandyoushould goshort. Themostcommonwaytotradethisstrategythen,istolookforacloseabovetheupper bandoraclosebelowthelowerband.Youthengolongorshortonthenextopen. Thisisastrategythatisfairlyeasytotestonhistoricaldata. Althoughtherearemanydifferentpermutationsofmarkets,timeframes,andposition sizing,thefollowingsimulationputsthissimplestrategytotestontheS&P500Index (SPX). TestOneRules: •WhenSPXclosesabovetheupperBollingerBandgolongonthenextopen •WhenSPXclosesbelowthelowerBollingerBandgoshortonthenextopen •Exitpositionby15%trailingstop

Settings: •Timeframe:Daily(EODquotes) •Start/Enddates:1/1/1990–1/1/2015 •Startingcapital:$100,000 •Positionsize:Fixed:$20,000pertrade •Commission:$0.01pertrade

TestOneResults: Torecap,thisstrategygoeslongwhenevertheS&P500Indexclosesaboveitsupper BollingerBandonthedailychartanddoessoonthenextopen.Itgoesshortonthe nextopenwhenevertheindexclosesbelowitslowerband. Thesystemrisksafixedamountof$20,000oneachtradeandexitsatradeby15% trailingstop.Asyoucansee,theresultsfromthissimplestrategyontheS&P500Index arelessthanimpressive:

jbmarwood.com



 Thereisacompoundedannualisedreturnofjust4.33%between1990and2015witha maximumdrawdownof30%.Thiscomparestoabuyandholdreturnof7.29%*per yearwithamaximumdrawdownof57%. Whenshortpositionsaretakenoutandthesystemisrunlongonly,theperformance improves,to5.41%CARwithadrawdownof22%. Unsurprisingly,theseresultsarenotspectacular,buttheyareactuallybetterthanIwas expecting.Differentpermutations,differentmarkets,timeframes,risk,mighthelpmake thisintoasimpleandworthwhilemodel. BollingerBandsTradingStrategyTwo jbmarwood.com

Asmentionedearlier,BollingerBandscanbeusedinmanywaysandyoucouldwritea wholebookjustaboutBollingerBandsstrategiesalone. Onesimplestrategycomesfromt hisarticleonthefinancesiteInvestopediaandit’s basicallythereverseofstrategyone. Theideaproposedinthearticleissimplytowaitforastocktoclosebelowthelower BollingerBandandthenbuyitthefollowingday.TheexactsellcriteriaisnotshownsoI willassumethatweexitthestockwhenthereverseoccurs.i.e.whenthestockcloses abovethetopBollingerBand.

 Accordingtothearticle,I’veincludedamaximumlossstoplossof30%. TotestthisstrategyIhavealsoincludedsomeextraliquidityrules(sothatweavoid thinlytradedpennystocks),arankingrule(inordertochoosebetweensignals)anda markettimingrule(sothatwedonottradeduringabearmarket).Inaddition,wewill spreadrisksothatwecanholdupto20positionsatonetime. ThesystemisrunonhistoricaldataforstocksintheS&P1500universebetween2000 and2015.Thisdataincludesdelistedtickersandisadjustedforcorporateactions. Herearethefullrulesandthenresults: TestTwoRules: •IfastockclosesbelowthelowerBollingerBand,buythefollowingdayontheopen •RankstocksbyRSI(20).Chooselowestrankedsignalsfirst jbmarwood.com

•Ifthestockclosesaboveit’supperBollingerBand,exitpositiononthenextopen •Maximumlossstoplossof30%

Settings/conditions: •Longonly •Liquidityrule:20dayaveragevolume>100,000 •Liquidityrule:Openprice>$2 •MarketTimingrule:SPXisaboveits80dayMA •Portfoliosize:Max20positions •Universe:S&P1500 •Timeframe:Daily(EODquotes) •Sample:1/1/2000–1/1/2015 •Startingcapital:$100,000 •Positionsize:5%pertrade. •Commission/slippage:$0.01pertrade

Asyoucanseefromtheresultsofthetest,thestrategywasabletoreturn17.25% between2000and2015withamaximumdrawdownof33%.



jbmarwood.com

 Thehighwinrateofthisstrategymakesitappealingandtheidealookstohave potential.Itmustbenoted,however,thattheresultsdodependongettinggoodprice fillsontheopen.Itestedthestrategyagainusingthebuypriceashalfwaybetweenthe openandthehigh,andthesellpriceashalfwaybetweentheopenandthelow,andthe annualisedreturndroppedto3.46%with42%drawdown. ToimprovethissystemIwouldsuggestlookingmorecloselyattheentrypriceand rankingmechanism. BollingerBandsTradingStrategyThree Goingbacktooursimulatorandlookingagainatour22rules,rulenumbersevenisthis: 7.Intrendingmarketspricecan,anddoes,walkuptheupperBollingerBandanddown thelowerBollingerBand. Thisisabittrickiertomodelusingthesimulator.Butitispossibletocomeupwitha strategybyspecifyingthenumberofhigheropensorclosesabovetheupperband.Or bythenumberofloweropensorclosesbelowthelowerband. ThisnextstrategygoeslongastockwhenitclosesabovetheupperBollingerBandfor twodaysinarow.ThisisaportfoliostrategyusingstocksfromtheS&P1500universe.

jbmarwood.com

ThestrategyusestheopenpricetocalculatethedefaultBollingerBandparameters (20,2)andtradesareenteredonthesamedayclose.

 Thesystemusesamarkettimingfiltersoonlytradeswhentheoverallmarket(SPX)is aboveit’s90dayMA.Anditusesaliquidityfiltertoavoidthinlytradedpennystocks andthelike. Thesystemisrunoninsampledatafirstthenoutofsample. TestThreeRules: •Buyastockonthecloseifitclosesaboveit’supperBollingerBand(calculatedwithopenprice)forthe seconddayinarow •RankstocksbyRSI(20).Chooselowestrankedsignalsfirst. •Exitpositionwith30%trailingstop

Settings/conditions: •Longonly •Liquidity:20dayaveragevolume>100,000 •Liquidity:Openprice>$2 •MarketTiming:SPXisaboveits80dayMA •Portfoliosize:Max20positions •Universe:S&P1500 •Timeframe:Daily(EODquotes) •InSample:1/1/2002–1/1/2012 •Outofsample:1/1/2012–1/1/2015 •Startingcapital:$100,000 •Positionsize:5%pertrade. •Commission:$0.01pertrade

jbmarwood.com

TestThreeResults: Theresultsshowa16.44%annualisedreturnwithamaximumdrawdownof30%during theinsampleperiodanda20.11%annualisedreturnwithamaximumdrawdownof 15%duringtheoutofsampleperiod.

 Essentially,thisisatrendfollowingstrategyanditshowsthestrengthofusingportfolios whentradingstocks. Oneofthereasonswhythestrategyworkswell,however,isthatitreliesontradingon theclose.Totradethissystemcorrectly,youwouldneedtoscanforpotential candidatesandbuyrightontheclose. ThiswouldnotonlyincludescanningbyBollingerBand–youneedtobesurethatthe stockisgoingtocloseabovetheupperbandwhichmaynotalwaysbeeasy–butalso byRSI,totakeaccountoftherankingcriteria.(Itestedthestrategyusingtheprior day’sRSIandtherewasverylittledifferenceinresultsothisisanotheroption). Ifwerunthetestagain,butthistimewebuyusingthehighpriceandnottheclose,the returngoesdownandthedrawdowngoesup,butnotbyagreatdeal.Thismightbea moreaccuraterepresentationofthepricefillswewouldactuallygetwhentradingthe system. Overall,theresultsareprettygood.Withsomedecenttoolsandalittlebitofdiscretion theremaybesomepotentialtobefoundhere.Belowisthefullequitycurvefrom 20002015:

jbmarwood.com

 BollingerBandsTradingStrategyFour Inrule6,JohnBollingerstatesthat: 6.Tagsofthebandsarejustthat,tagsnotsignals.AtagoftheupperBollingerBandis NOTinandofitselfasellsignal.AtagofthelowerBollingerBandisNOT inandofitselfabuysignal. Butwhatifthosetagsoccurwhenthestockisalreadyinanuptrend?Maybethereisa possibilitythataclosebelowthebottomBollingerBandcouldactuallybeagoodtimeto tradethepullback? Inthisnextstrategy,weusethesamesettingsastesttwoexceptfortwodifferences. Weneedthestocktobeinanuptrendandweonlyexitbytrailingstop.

jbmarwood.com

 TestFourRules: •BuyastockonthecloseifitclosesbelowitslowerBollingerBand •RankstocksbyRSI(20).Chooselowestrankedsignalsfirst •20daymovingaveragemustbehigherthanthedaybefore •Exitpositionwith30%trailingstop

Settings/conditions: •Longonly •Liquidity:20dayaveragevolume>100,000 •Liquidity:Openprice>$2 •MarketTiming:SPXisaboveits80dayMA •Portfoliosize:Max20positions •Universe:S&P1500 •Timeframe:Daily(EODquotes) •InSample:1/1/2002–1/1/2012 •Outofsample:1/1/2012–1/1/2015 •Startingcapital:$100,000 •Positionsize:5%pertrade. •Commission:$0.01pertrade

TestFourResults: Ascanbeseenbelow,theresultsareOK.Annualisedreturnintheinsampleperiod was14.35%withadrawdownof28%.Intheoutofsampleperiod,theannualised returnwas15.88%withadrawdownof15%.

jbmarwood.com

 Again,thestrategysuffersfromtheproblemoftradingrightontheclose,sothis strategy(andstrategythree)mightbenefitfromsomediscretion. Hereisthefullequitycurvebetween2000and2015:

 Conclusion

jbmarwood.com

Overall,itappearsthattradingstrategythreeisthemostappealing,basedonthe CAR/MDDratio.However,thehigherwinrateofstrategy2suggeststhatstrategy2 mightalsobeworthinvestigating. Asmentioned,theproblemofscanningfortherightsignalsandtradingthemrighton theclosemeansthatstrategythreeandfourarerelativelydifficulttotrade,atleast withoutmoreexpensivetools.Whereastradingstrategytworeliesongettinggoodprice fillsontheopen. Nevertheless,thereappearstobepotentialinthesesystemsforfurtherdevelopment.  Strategy3SystemCode: //////////////////////////////////////////jbmarwood.com //

BollingerBands

//

////////////////////////////////////////// SetFormulaName("bbands"); size1=20;//Optimize("size1",18,18,34,2); SetOption("InitialEquity",100000); SetOption("MinShares",1); SetOption("MarginRequirement",100); SetOption("UsePrevBarEquityForPosSizing",True); SetTradeDelays(0,0,0,0); SetOption("MaxOpenpositions",size1); SetOption("AllowSameBarExit",false); Numberpositions=size1; SetOption("Maxopenpositions",numberpositions); SetPositionSize(1,spsShares); PositionSize=100/size1; PositionScore=100/rsi(20);//ranking //*Amendmentfordelistedsecurities Delisting=GetFnData("DelistingDate"); ThisIsLastBar=BarIndex()==LastValue(BarIndex());

jbmarwood.com

ExitLastBar=datetime()>=GetFnData("DelistingDate"); Cond=IIf(IsNull(delisting),1,DateTime()0;//*fixsituationwherestocktradesonfirstbar Liquid=O>2ANDOpenInt>2ANDMA(V,20)>100000ANDfix; //*Markettiming spx=Foreign("$SPX","o"); level=80; bullmarket=spx>MA(SPX,level); //*TradingSystem bbandup=c>BBandtop(O,20,2)ANDRef(c,1)>Ref(BBandtop(O,20,2),1); Buy=bbandupandliquidANDcondANDbullmarket; Sell=0ORthisislastbarORexitlastbar; ApplyStop(stopTypeTrailing,stopModePercent,trailpercent); Equity(1); BuyPrice=c;//(o+H)/2; SellPrice=c;//(o+L)/2;  Buy=ExRem(Buy,Sell);//*removesextrasignals Sell=ExRem(Sell,Buy); ///////////////////////////////////////////////////////// /////////////////////////end///////////////////////////

 

jbmarwood.com...