Index

Sr No 1 2 3 4 5 6 7 8 9 9.1 9.2 9.3 9.4 9.5 Topic IntroductiontoAlgorithmTrading HighFrequencyTrading AlgorithmTradingVsHighFrequencyTrading AlgorithmTradinginAsiaPacific EvolutionofElectronicTradinginIndia SmartOrderRouting,DMAandColocation CurrentinfrastructureinIndia CurrentTradeStructureinUS&India HFTanditsimpactonMarketQuality DataSample HFTactivityinUS DoHFTsfleeinvolatilemarkets? DoHFTscontributetoPriceDiscoveryProcess? WhatroleHFTplaysinprovidingmarketliquidity? 2 9 11 12 14 15 18 20 23 23 24 25 27 28 31 33 39 43 45 49 54 PageNumber

9.6 DoHFTgenerateordampenvolatility? 10 PossibleManipulationsUsingHFT/AlgorithmicTrading 11 12 13 14 15 CurrentRiskManagementframeworkinIndia Algoapproval,registrationandriskmanagementframework Concerns Recommendations Conclusion

Annexure
Sr Topic No 1 RiskmanagementcriteriasforapprovalofDMAfacilityatNSE 2 ProcedureforgrantingpermissiontoTradingMemberforAlgorithm 3
4 Trading CurrentRiskManagementframeworkinIndia AlgorithmicandHighFrequencyStrategies Executionstrategies

PageNumber 55 57 58
66

70

StructuralIssuesConcerningAlgorithmTradinginIndia

Chapter1

IntroductiontoAlgorithmTrading

Definitionofalgorithmtrading
Algorithmtradingreferstoautomatedtradingbasedonalgorithm.Thepurposeofusing automatictradingistoanalyzepriceandmarketconditionsinordertotradeatthe minimumcostbasedontherelevantalgorithmmethod. Algorithmic Trading can be defined as the use of computer programs for entering trading orderswherethecomputeralgorithmdecidesonaspectsofthetradeexecutionsuchasthe timing, price, or quantity of the order. Algorithm Trading is also known as automated trading, algo trading, blackbox trading or robo trading. Algorithms dynamically monitor market conditions across different securities and trading venues to make the trade executiondecision.Algorithmictrading(AT)ismorecomplexthanelectronictradingandit encompassesmanyformofcomputeraidedtradingincludingHighFrequencyTrading. Algorithmic Trading is widely used by pension funds, mutual funds, and other buy side (investor driven) institutional traders, to divide large trades into several smaller trades in ordertomanagemarketimpact,andrisk.Infact,algorithmswereoriginallydevelopedfor use by the buyside to manage orders and to reduce market impact by optimising trade executiononcethebuyandselldecisionshadbeenmadeelsewhere.Sellsidetraders,such asmarketmakersandsomehedgefunds,provideliquiditytothemarketbygeneratingand executingordersautomaticallywiththehelpofalgorithms.

Algorithmtradingfocusesonminimizingthemarketimpactbysplittingtrades.Algorithm tradingwasoriginallydevelopedasanorderplacementsystemforthepurposeof minimizingtradingcost,butitisnowbeingusedasanoveralltermtodescribeitsstrategy andprocess.

StructuralIssuesConcerningAlgorithmTradinginIndia

TypesofAlgorithmTrading
AlgorithmTradingisbroadlycategorisedintoHighfrequencytrading,BasketTrading,Multi ExchangetradingandSituationalAlgorithmTradingcategory.BasketTrading(Program trading)isgenerallyusedforstockbaskettrading.Multiexchangetradingistradinga productacrossdifferentexchangesusingcomputeralgorithm.Multiexchangetradingis facilitatedbySmartOrderRoutingalgorithms. Smartorderroutingisatechniqueofusingtheclosecorrelationofmajorexchangesand crosslistingofacompanyi.e.,placingordersonanexchangewherethemostfavourable conditionsareseen.InIndia,SEBIalloweduseofSmartOrderroutinginAugust2010and currentlyitisoperationalinbothCapitalMarketaswellasCurrencyDerivativeSegment. FlashTradingreferstopracticeinUStoflashordersforfewsecondstoselectmembers beforesubmittingittoordermatchingplatform.Situationalalgorithmtradingcategory hereafterreferredasAlgorithmTrading/AlgorithmicTrading/ATinterchangeably,refersto anyandallkindofcomputeraidedtradingwhichdoesnotfallunderfirstthreecategories. Situationalalgorithmdetermineswhattotrade,whentotradeandhowtotrade

AlgorithmTrading

HighFrequency Trading

BasketTrading (Programm Trading)

MultiExchange Trading (Tradingusing SOR)

FlashTrading

Situational Algorithmic Trading

Strategiesoralgorithmsusedbythecomputertoautomate(enter,modifyorcancel)trade ordersarecalledalgorithmictradingstrategiesandtheyarebroadlyclassifiedasexecution strategies,profitseekingstrategies(Alphastrategies)andarbitragestrategies. Arbitragestrategies Thesimultaneouspurchaseandsaleofanassetinordertoprofitfromadifferencein theprice.Itisatradethatprofitsbyexploitingpricedifferencesofidenticalorsimilar financialinstruments,ondifferentmarketsorindifferentforms.Arbitrageexistsasa resultofmarketinefficiencies;itprovidesamechanismtoensurepricesdonotdeviate substantiallyfromfairvalueforlongperiodsoftime.

StructuralIssuesConcerningAlgorithmTradinginIndia

ProfitSeekingstrategies(Alphastrategies) Alphaisacommonwayofmeasuringtheperformanceofatradingstrategyinterms ofriskadjustedreturninexcessofabenchmarkindexora"riskfree"investment. Statisticallydrivenstrategiesareutilizedtogeneratestocktradingsignalsonboth longandshortsideswiththegoalofachievingconsistentaboveaveragereturns andapositiveAlpha. AlphaStockStrategiesmayperiodicallycalculateandpublishAlphasrelativetothe market(S&P500index)orotherappropriatebenchmarks,suchasanaggregate long/shorthedgefundindex. Executionstrategies Algorithmtradingiswidelyusedamonginstitutionalinvestorslikepensionfundsand investmentcompaniesbecausecostcanbesavedthroughbulkorders,reducingthe impactonthemarketandpreventingtheexposureoftradinginformation.Algorithm tradingalsohelpsreducetradingcost. ThemostcommontypesofexecutionalgorithmareVWAPandTWAP.VWAPsplitsorder volumebasedonhistoricaldata.TWAPsplitsorderovertime.Themarketparticipation strategyismanipulatingtradingvolumesothatspecificordersdonotaccountfora significantpartoftotaltradingvolumeonthemarket.Ontheotherhand,theinline strategyismanipulatingordersorpricessothattheydonotsurpassalimitprice.

StructuralIssuesConcerningAlgorithmTradinginIndia

Algorith hmTrading gStrategie es

Arbitr ragestrat tegies

AlphaS Strategies s
Directional Trading Strategies Mean Revesion es Strategie

CostRedu C uction/ Exe ecutionS Strategies

Benc chmark algo orithms (VWAP, Imple ementati onSh hortfall, Partic cipation (%V Volume) ,Targe etClose, TW WAP, InLin ne,etc) Liquidity seeking algorithms s (CrossFire, , Hunt, c) Iceberg,etc

Delta a Nuetr ral

Sta atistical Arb bitrage

EventBa ased Arbitra age

Scalping Strategies

Dark Liquidity g Seeking Strategie es

Detailsabout tTradingstr rategiesareoutlinedinA Annexure4 &Detailsab bout Note:(D Benchma arkalgorithm msandLiquidityseeking galgorithmsareprovidedinAnnexure5)

StructuralIs ssuesConcer rningAlgorith hmTradingin nIndia

Reasonsforusingalgorithmsintrading
InasurveyconductedbyTRADEgroup,22%participantssaidtheypreferalgotrading becauseitprovidesthemanonymityoftradeavoidingleakageofinformationwhichpossibly mayleadtofrontrunning.20%saidcostreductionintradeexecutionisthemainpurpose whytheyusealgotrading.14%participantssaid,algotradingallowstheirtradertofocuson tradingstrategythantradingexecution,sothattradesproductivitycouldbeincreased.13% preferalgoasitreducesmarketimpact.11%peoplelikealgobecauseofitsspeedof computationandexecution.7%preferalgosbecauseitiseasytouseand4%preferitdue tocustomisationfeaturesofferedbyalgos.

StructuralIssuesConcerningAlgorithmTradinginIndia

TypesofAlgorithmsUsedbythebuysidefirms

Source:AlgorithmTradingSurvey2010conductedbyTradingscreen Overtheyears,algousagehasevolvedfromlesssophisticatedparticipationstrategiessuch asVWAPandTWAPtomorecomplexpriceimprovementapproachesthatseekto minimiseslippagefromatargetprice.Asaresultin2010,Implementationshortfall(IS)for singlestockshasriseninpopularity,upfrom39%in2008to68%.VWAPsawthefirst notabledeclinesince2008initsuseamongbuysiderespondentsdownfrom64%to58%. UseofTWAPalsodeclinedmarginallyfrom23%in2010to20%in2009. Darkliquidityseekingalgorithmssearchesallthedarkpoolseliminatingtheneedtosplitan orderinsearchofliquidity.Darkliquidityseekingalgorithmsaremostwidelyused algorithmsbybuysidefirmsin2010,constituting81%ofrespondentsuse,upfrom51%a yearago.Similarly,albeitfromalowbase,therehasbeenafivefoldincreaseintheuseof algorithmstotakeadvantageofinternalcrossingopportunities,upfrom5%in2009to25% in2010.Percentageofrespondentsusingvolumeparticipationstrategyandbasket implementationshortfallstandat62%and20%respectively.

StructuralIssuesConcerningAlgorithmTradinginIndia

USEquityShareVolumebyMarketParticipant
InUS,HFTconstitute61%oftotalvolumetraded(28%ofthemweremarketmakersand 29%ofthemwereIndependentHFTs).InnonHFTcategory,Investmentbanksand proprietarytradingis16%.Retailparticipationislimitedtoonly3%ofvolumetraded.
US Equity Share Volume by Market Participant2009
28% 29%

13% 4% 7% 3%

16%

Independ ent HFT

Retail

Investme nt Bank Prop

HFT Broker/ Market Makers

Hedge Funds

HFT Hedge Funds

Long Only

HFTs(MarketMakers/LiquidityProviders/SellSidefirms) NonHFTs(Investors/BuySidefirms) SourceTABBGroup 39%volumeinUS 61%volumeinUS

StructuralIssuesConcerningAlgorithmTradinginIndia

Chapter2

HighFrequencyTrading
Highfrequencytrading(HFT)isasubsetofalgorithmictradingwherealargenumberof orders(whichareusuallyfairlysmallinsize)aresentintothemarketathighspeed,with roundtripexecutiontimesmeasuredinmicroseconds.Programsrunningonhighspeed computersanalysemassiveamountsofmarketdata,usingsophisticatedalgorithmsto exploittradingopportunitiesthatmayopenupformillisecondsorseconds.Participantsare constantlytakingadvantageofverysmallpriceimbalances;bydoingthatatahighrateof recurrence,theyareabletogeneratesizeableprofits. Typically,ahighfrequencytraderwouldnotholdapositionopenformorethanafew seconds.EmpiricalevidencerevealsthattheaverageU.S.stockisheldfor22seconds. TABBgroup,afinancialservicesindustryresearchfirm,estimatesthatHFTnowaccountsfor 56%ofturnoverinUSAand38%turnoverinEuropeintheequitysegment.

WhatareBuysidefirmsandSellsidefirms?
Buysidefirmsrefertoinstitutionsconcernedwithbuying,ratherthanselling,assetsor securities.Privateequityfunds,mutualfunds,unittrusts,hedgefunds,pensionfunds,and proprietarytradingdesksarethemostcommontypesofbuysideentities. Buysidefirmsusuallytakespeculativepositionsormakerelativevaluetrades.Buyside firmsparticipateinasmallernumberofoveralltransactions,andaimtoprofitfrommarket movements.

StructuralIssuesConcerningAlgorithmTradinginIndia

Sellsidefirmsreferstoinstitutionsthattakeordersfrombuysidefirmsandthen"work" theordersbyslicingthenintosmallorders.BrokeragefirmsandMarketmakingfirmsare mostcommonexamplesofsellsidefirms. Sellsidebrokeragesareregisteredmembersofastockexchange,andsometimestheyare requiredtobemarketmakersinagivensecurity. SellsidefirmsprofitfromCommission(brokerage),advisoryfeechargedtobuysidefirm andthebidofferspread.

Widelyusedhighfrequencytradingstrategies

10

StructuralIssuesConcerningAlgorithmTradinginIndia

Chapter3

AlgorithmTradingVsHighFrequencyTrading
AlgorithmicTrading ATistheuseofcomputerprogramsforentering trading orders with the computer algorithm deciding on aspects of the order such as the timing,price,orquantityoftheorder Algorithmic Trading is computer aided but not necessarilyHighfrequencyorlowlatencytrading Accounts for 70% volume traded in US equity in 2010.NonHighfrequencyAlgorithmicTradingis only14%inUS. ATfirmsholdingperioddependsonstrategythey employ.Theymayholdpositionsforfewminutes or for few years or they might never sell it. Typicallyaverageholdingperiodislong. HighFrequencyTrading HFT is highly quantitative, employing computerized algorithms to analyze incoming market data and implement proprietary trading strategies HFTispartofAlgorithmicTrading Accounts for 56% volume traded in US Equity in 2010. HFT usually implies a firm holds an investment positiononlyforverybriefperiodsoftimeeven justsecondsandrapidlytradesintoandoutof thosepositions

AT firms include both DayTraders and Position HFT firms are typically DayTraders and end a Traders and may carry overnight position trading day with no net investment position in dependingonstrategytheyuse thesecuritiestheytrade. AToperationsareusuallyfoundinBuySidefirms (such as Mutual funds, Insurance companies, Pension funds, FII, Investment Banks, some HedgefundsandsomeProprietarytradingfirms) Selecting High Profit Making and properly backtested algorithms is more important than speed.SpeeddoesmatterinATtoo,butneedfor speedisnotanutmostpriority. ApplicationofAlgoTrading Directional Trading : Trend Following, Pair Trading,MeanReversion,Scalping Arbitrage : Delta Neutral Strategies, Statistical Arbitrage Transaction cost reduction(Trade Execution) : VWAP, TWAP, Liquidity seeking, Implementation shortfall HFT operations are usually found in proprietary firms or on proprietary trading desks in larger, diversifiedfirms HFT strategies are usually very sensitive to the processingspeedofmarkets(alsocalledLatency) andoftheirownaccesstothemarket. ApplicationofHFT RebateTrading,MarketMaking, FilterTrading,MomentumTrading, StatisticalArbitrageandTechnicalTrading

11

StructuralIssuesConcerningAlgorithmTradinginIndia

Chapter4

AlgorithmTradinginAsiaPacific
AlgorithmicTradinginAsiaPacific Amongasiapacificcountries,algorithmicactivityishighestinHongKong(39%)followedby Singapore(31%),Japan(26%),Australia(20%)andIndia(15%).Algorithmictradingin HongKongisslatedtoreach61%in2012fromcurrent39%.Indiawascautiousinitsuseof AlgoTrading,buthaslotofpotentialtogrow.Bytheendof2012,ATactivityinIndiais expectedat30%fromcurrent12%.

Source:CelenetResearch Asiaisaveryexcitingmarketforelectronictradingofcashequityandderivatives.Whilethe levelofdevelopmentofeachmarketdiffers,theleadingexchangesinJapan,Australia, Singapore,HongKong,andIndiashouldsetthetone,withfastermatchingengines, enhancedmarketdata,andcolocationservices.Withahighproportionofitsdomesticbuy sidebeginningtousealgorithmictrading,HongKongisbestpoisedtogrowintheregion.It isfollowedbySingapore,wheretheregulatoryenvironmentisveryconducivetoalgo trading.Similarly,Japan,andAustraliaareexpectedtohavehigheralgotradingincoming years.Finally,Indiahasbeenaslowstarterbutisrapidlycatchingupwithsomeofits counterpartsandisexpectedtohavealgotradinglevelsofaround30%by2012 AnshumanJaswal,CelentSeniorAnalyst

12

StructuralIssuesConcerningAlgorithmTradinginIndia

 OnMarketEnvironmentFront Singapore&HongKongScoreTopinchart India Japan Australia Taiwan China OnRegulatoryForcesFront HongKong Australia Singapore Japan Taiwan India China(MostRestrictedmarketinworldfortrading) OnTechnologyEnablersFront Singapore(HighestinvestmentontechnologyfrontinAsia) Australia Japan HongKong India Taiwan China OnAlgorithmicTradingFront HongKong(48%ofvolumetradedalgorithmically) Singapore Australia India&Japan China

13

StructuralIssuesConcerningAlgorithmTradinginIndia

Chapter5

EvolutionofElectronicTradinginIndia
MilestonesinIndianEquityMarkets

DomesticBuySide startedAlgoTradingoperations SmartOrderRoutingIntroduced (DomesticSellSide startedAlgoTradingoperations) DMAIntroduced AlgorithmicTradingBegins IntroductionofDerivativesinIndia InceptionofNSEandStartofelectronicTrading

1994'

200001

200304

2008'

2010'

201011

2011'

Inspiteofthehiccups,theIndianmarketissetforsomeinterestingtimesahead.Asthe issuesbetweentheexchangesareresolvedandregulatorsbecomemoreadeptathandling therapidadvancesintradingtechnology,theIndianexchangesandbrokerageswillcompete fordominanceintheAsiaPacificregion.saysAnshumanJaswal,CelentSeniorAnalystand authorofthereport."

14

StructuralIssuesConcerningAlgorithmTradinginIndia

Chapter6

SmartOrderRouting,DMAandColocation

SmartOrderRouting
InAug2010SEBIalloweduseofSORinIndia,BSEintroducedSORfacilityfortrading membersonSept29,2010andNSElauncheditonOctober13,2010.SORfacilityisallowed tobeofferedtoallclassofinvestorsasdirectedbySEBI.Membersdesirousofproviding SORfacilityshallmakeanapplicationtotheExchangeinspecifiedformat. Membersarerequiredtotestthesoftwareonassociatedexchangestradingtest environmentduringaprespecifiedtime.Onsatisfactorycompletionoftestingonthe Exchangetestenvironment,thememberisrequiredtogivecomprehensivedemonstration oftheirSORfacilitytotheExchange.Onfulfilmentoftheconditionsassatisfactoryand meetingSEBI/Exchangeminimumrequirements,theExchangewouldgrantpermissionto themembertocommenceSORfacility. Requirements: a. SORcanbeusedonlytorouteorderstorecognisedstockexchanges b. ThebrokerserverroutingordersplacedthroughSmartOrderRoutingsystemofthe ExchangeshallbelocatedinIndia. c. Thetradingmembershalldeterminetherecognisedstockexchangewheretoroute theorderorpartoftheorderbasedonfactorslikeprice,costs,speed,likelihoodof executionandsettlementsizeandnaturerelevanttotheexecutionoftheorder. d. IdentificationofordersroutedthroughSmartOrderRouting:13thdigitof15digit ordercodeisusedtoidentifySORorderatNSE.

AtBSE14thDigitof15digitorderisusedtoidentifySORorder 14thdigit 4or6 "4"OnlySOR "6"SORthroughDMA e. ThesoftwareandsystemsproposedforSORisrequiredtobedulycertifiedby SystemAuditorbeforegrantofpermission.

15

StructuralIssuesConcerningAlgorithmTradinginIndia

DirectMarketAccessFacility
DirectMarketAccess(DMA)isafacilitywhichallowsbrokerstoofferclientsdirectaccessto theexchangetradingsystemthroughthebrokersinfrastructurewithoutmanual interventionbythebroker. SomeoftheadvantagesofferedbyDMAaredirectcontrolofclientsoverorders,faster executionofclientorders,reducedriskoferrorsassociatedwithmanualorderentry, greatertransparency,increasedliquidity,lowerimpactcostsforlargeorders,betteraudit trailsandbetteruseofhedgingandarbitrageopportunitiesthroughtheuseofdecision supporttools/algorithmsfortrading. SEBIallowedTradingMemberstoprovideDMAfacilityinApril2008fortradingin capitalmarketsegment.InJune2008,NSElauncheditsDMAfacilityforitsmembers. Memberscanusetheirownsoftwarerunningonanysuitablehardware/software platformoftheirchoiceandhencecustomizethesoftwaretomeettheirspecialised needssuchasprovisionofonlinetradeanalysis,riskmanagementtools,integration ofbackofficeoperationsetc BostonbasedresearchandanalysisfirmCelentestimatesthattheDMAintheUS accountsforabout20%ofequitiessharevolumein2010.InEurope,DMAflowfor equitiesisexpectedtogrowfrom8%oftradedvalueto15%in2011,accordingto thefirm. Celnetestimates,in201011%oftradedvolumeinIndiawasthroughDMAfacility. Theinstitutionalinvestors,whichcurrentlyaccountfor2225%ofthetotalturnover intheIndianstockmarkets,areexpectedtoincreasetheirmarketshareofthe turnoverwiththeadventofDMA. ThebasicbrokeragefeesforinstitutionaltradesinIndiarangefrom10to40basis points,Marketanalystsexpectthisfeetodropto10to20basispointsfollowingthe adoptionofDMAfacility. Majorplayers:MFGlobal,BNPParibas,MorganStanley,UBSandCreditSuisse BrokerprovidingATaswellasDMAservices:MerrillLynch,GoldmanSachs, JPMorgan,CitigroupandCreditSuisse Requirements The software and systems proposed for DMA shall be duly certified by the Exchange empanelled System Auditor before grant of permission. TradingmembersprovidingDirectMarketAccessfacilityarerequiredtopopulate thefollowingvaluesinthefieldprovided(15digits)intheorderstructureforevery orderemanatingfromDMAterminals.

16

StructuralIssuesConcerningAlgorithmTradinginIndia

Member seeking an permission should comply with risk controls checks as specifiedinannexureI

Colocation
Colocationisthepracticeoflocatingbrokerorclienttradingsoftwareandhardwarein closeproximitytothetradingplatformstradingengine.Theaimistominimisepropagation andtransmissionlatency.Exchangesandthirdpartyproviderschargeafeeforofferingco location. IndianExchangeshavestartedprovidingcolocationfacilitytotheirmemberbrokers, wherebytheycanplacetheirtradingserversclosetotheexchangesengineonafirstcome firstservedbasis.Colocationsavescrucialmillisecondsfromthetimeittakestoplacean orderanditsreceiptattheotherend.Thebrokerwithhisservernexttotheexchange enginegetsapricefeedthatisupdatedeverythreefourmilliseconds,whileabrokerata remoteplacewillgetthisfeedupdatedevery3040milliseconds. InFuturesIndustryAssociation(FIA)conferenceinMumbai,anofficialfromtheNational StockExchange(NSE)said60%oftheorderscomingintotheexchangewerefromco locatedservers.

17

StructuralIssuesConcerningAlgorithmTradinginIndia

Chapter7

Curre entinfras structure einIndia

ivity Connecti Indianbr rokersneedFIXconnectivitytoattra actinternationalorderflowandmanyhave madesig gnificantinve estmentsinFIXnetwork ksandordermanagementsystems( (OMSes).The widespre eadadoption noftheFIXp protocolove erthelasttw woyears,allowedmanycountriesto o allowDM MAfacility. bothNSEand dBSEhaveF FIXconnectiv vity. InIndiab Latency Latencyr referstoamountoftime etakentoel lectronically ysendorrec ceiveinforma ation.HFTs relyonverylowlatency(i.every yquickdatat transmission n)fortradingstrategies. .Worldwide e esareinvest tinginnewt technologyt toreducelat tency. exchange NYSETec chnologiesU UniversalTra adingPlatfo orm 3milliseconds TimeLag gfordatafee ed TimeLag gfororder& &cancel Acknowle edgment TimeLag gforRoundT TripExecutio on at(timefromorderconfirmationtotrade onfirmation) ) executiontotradeco 2milliseconds Nasdaqlistedissue 650micros seconds NYSElist tedissue9 950microsec conds BATS:40 00microseco onds DirectEd dge:300500 0microsecon nds

Recently, ,NYSElaunc chedEtherne etbasedma arketdatad deliverysolu utionthatha as25 microsec condslatencyatratesof f1million20 00bytemessagesperse econdperco ore. Ordersp peed(Round dTripExecut tionTime)atmajorexch hanges

KRX K 80, ,000 S

HK KE 10,0 000 S

SGX 000 10,0 S S

BSE E 10,00 00 S

LSE 0 6,000 S

TSE 5,000 S

NSE 5,000 S

ASX 250 S

NDAQ 143 S

NYSE 25 S

18

StructuralIs ssuesConcer rningAlgorith hmTradingin nIndia

Capacity Capacityreferstotheamountofdatathatcanbesimultaneouslyelectronicallysentor received.Itisanecessaryprerequisiteformanyalgorithmictradersthatatradingsystem canreliablysustainlargevolumesofdata.Forexample,electronicmarketmakersconstantly checkmarketpricesandadjusttheirquotesaccordingly.Statisticalarbitrageursare constantlylookingformispricingopportunities. Bothstrategiesrequiretheabilitytosend andreceivelargeamountsofdatasimultaneously,likerequestingcurrentbids/offers, sendingtheirownbid/offerandthenreceivingconfirmationthatthetradingenginehas receivedtheirorderandanysubsequentamendmentorcancellationofthatorder. WithintroductionofAlgorithmTradingnowmanystockexchangesarestrugglingtokeep pacewitheverincreasingcapacitydemands.Theyareinvestinginnewtechnologiesto increasecapacityinordertofacilitateMultilateralTradingFacility. Inresponsetochangingenvironment,BSEincreasedcapacityfrom1000orderspersecond to20,000orderspersecond.NSEalsoisworkingtowardsincreasingcapacity.

19

StructuralIssuesConcerningAlgorithmTradinginIndia

Chapter8

CurrentTradeStructureinUS&India

AverageTradesizeandNumberoftradesinUS
Averagesharespertrade(AverageTradesize)decreasedby83%from1,477sharesin1997 to244sharesin2009,whileAveragetradesperDay(NumberofTrades)increasedby26 timesfrom743tradesin1997to19,943tradesin2009.AveragesharesPerDayincreased from1.09mn(1477*743)in1997to4.86mn(19943*244)in2009. Thefiguresshowthatthereisbeenincreaseinnumberoftradesexecutedanddecreasein ordersize.Thiscanbeattributedtointroductionofalgotradingwhichslicelargeorderinto manysmallorders.Thenegativeimpactofthisisthatitmayfloodthesystemwithlarge numberofordersincreasingtheloadonexchangestradingsystem.Alsoitincreasesclearing andsettlementcostsfortraders. US Equity Share volume and trades 40,000 1,600 35,000 1,400 30,000 1,200 25,000 1,000 20,000 800 15,000 600 10,000 400 5,000 200 0 0 Avg Shares Per Day (mm) Avg Trades Per Day Shares / Trade Source:TABBGroup Avg. Shares / Trade

20

StructuralIssuesConcerningAlgorithmTradinginIndia

AverageTradesizeandNumberoftradesinIndia
In2010,NSEwasthe4thlargestexchangeintheworldintermsofnumberoftrades executedinequitysegmentfollowedbyNYSE,NASDAQandSSE.BSEwas7thlargest exchangeovertakingmanyEuropeanandAsianexchanges.

NumberofTradesinEquitySegment(inCrore)
205 174 166 156 133 92 56

NYSE NASDAQ Shanghai National Shenzhen Korea BombaySE Euronext OMX SE Stock SE Exchange (US) Exchange India

Source:wfe
Buttheinterestingpointhereisthatintermsofleastnumberofsharespertrade,Indiaisalsothe firstcountryintheworldwithjust120sharestradedpertradefollowedbyfollowedbyNASDAQ (230),KoreaExchange(278)andNYSE(293).

SharesPerTradeinEquitySegment
1,559 1,218

293

230

120 ShanghaiSE National ShenzhenSE Stock Exchange India

278

NYSE Euronext(US)

NASDAQ OMX

Korea Exchange

Source:wfe

21

StructuralIssuesConcerningAlgorithmTradinginIndia

AverageOrderCancellationRate
SinceintroductionofAT,OrdertoTraderatiohassignificantlygoneupinmanycountries includingIndia.AtNYSEordertotraderatiois30:1,whichmeansonlyoneorder,outof30 ordersentered,getsexecutedand29othersareeithermodifiedorcancelled.ATBSEorder totraderatiois19:1.

Theincreaseintheaverageordertotraderatioandaverageordercancellationratemayno betheconcernaslongasexchangeshavebuiltsufficientlylargecapacityandhavelimited orderentryratethroughalgorithms.

Commission,ClearingandSettlementchargesandTransactioncost
ItisbeenobservedworldwidewithintroductionofATandDMAfacilitytherehasbeen considerabledecreaseinthecommissionchargedasmanyinstitutionsprefertoexecute tradesdirectlyusingDMAfacility.AlsowithadventofATandHFT,therehasbeenan increaseinthenumberofordersexecuted,whichcausesclientstonegotiatecommissionin percentagetermwiththeirbrokers.Forinstitutionbrokerageis1040bps,whereasfor retailinvestorsitisintherangeof2050bps. ClearingandsettlementchargesareskyhighinIndiaespeciallyforretailinvestors.These chargesareapplicableonlyonselltransactionsandvariesfrom1030Rsforevery transactionfrombrokertobroker.Irrespectiveofnumberofsharessold,youhavetopay clearingandsettlementcharges,whichIbelievearequithighforretailinvestorsselling10 100shares.

TransactioncostforintradaytransactionisreasonablebutforDeliverytransactionitisquite highataround0.14%ofturnoverforproprietarytraders,and0.5%forretailinvestors.

22

StructuralIssuesConcerningAlgorithmTradinginIndia

Chapter9

HFTanditsimpactonMarketQuality
9.1DataSample
DatasamplechosenforstudyingimpactofHFTinstockslistedoneitherNASDAQorNYSEis asfollows Sample:Listof120stockslistedoneitherNASDAQorNYSE AA ARCC BZ CNQR CTSH FFIC GPS KTII MOD PNY AAPL AXP CB COO DCOM FL HON LANC MOS PPD ABD AYI CBEY COST DELL FMER HPQ LECO MRTN PTP ADBE AZZ CBT CPSI DIS FPO IMGN LPNT MXWL RIGL AGN BARE CBZ CPWR DK FRED INTC LSTR NC ROC AINV BAS CCO CR DOW FULT IPAR MAKO NSR ROCK AMAT BHI CDR CRI EBAY GAS ISIL MANT NUS ROG AMED BIIB CELG CRVL EBF GE ISRG MDCO NXTM RVI AMGN BRCM CETV CSCO ERIE GENZ JKHY MELI PBH SF AMZN BRE CHTT CSE ESRX GILD KMB MFB PFE SFG ANGO BW CKH CSL EWBC GLW KNOL MIG PG SJW APOG BXS CMCSA CTRN FCN GOOG KR MMM PNC SWN Source:HighFrequencyTradingandItsImpactonMarketQualitybyJonathanBrogaard, Sept2010 Note:Thesestatisticsareanaggregatefor26HFTfirmstradingactivityin120stocksfor fiscalyearendingonDecember31,2009.

23

StructuralIssuesConcerningAlgorithmTradinginIndia

9.2HFTactivityinUS
HFTsareinvolvedin68.5%ofalldollarvolumetradedinthesample.Theirlevelofdaily involvementvariesfrom60.4%to75.9%.Theydemandliquidityin42.7%ofalldollar volumetradedandsupplyitin41.1%. HFTsparticipatein73.8%ofalltradesandparticipationvariesatthedaylevelfrom65.1%to 81.9%.Theydemandliquidityin43.6%ofalltradesandsupplyitin48.7%. PanelA>DollarVolumetradedbyHFTsaspercentageofTotalDollarVolumetradedinthe sampleof120stocks PanelB>QuantitytradedbyHFTsaspercentageofTotalQuantitytradedinthesampleof 120stocks Table1

Source:HighFrequencyTradingandItsImpactonMarketQualitybyJonathanBrogaard, Sept2010 Note:Thesestatisticsareanaggregatefor26HFTfirmstradingactivityin120stocksfor fiscalyearendingonDecember31,2009.

24

StructuralIssuesConcerningAlgorithmTradinginIndia

9.3DoHFTsfleeinvolatilemarkets?
WhileHFTmaybeincreasinglyprovidingliquiditytothemarketintheplaceofmore traditionalmarketmakingactivities,someinvestorssuggestedthatunlikeregisteredmarket makersontradingvenuesthereisnoobligationorincentiveforhighfrequencytradersto continuetoprovideliquiditytothemarketintheeventofadversemarketconditions(high volatility). ResearchshowsasindicatedinGraph1,HFTsactivityappearstobealmostflatacross volatilitylevels.EvenonthemostvolatiledaysHFToverallactivitydoesnotseemto increaseordecreasesubstantially.However,whenvolatilityislowHFTsactivityislessthan average. Graph1 HFTactivitymorethanaverage

HFTactivityLessthanaverage

Source:HighFrequencyTradingandItsImpactonMarketQualitybyJonathanBrogaard, Sept2010 Where, HFT%ChangePercentchangeinHFTactivity HFT%MA(9)9DaymovingaverageofpercentchangeinHFTactivity 95%confMA(9)UpperandLowerpricebandofHFT%MA(9) HFTprovideabout10%moreliquiditythanusualonverylowvolatilitydays.ThelevelofHFT liquidityslowlydeclinesasvolatilitypicksup,atthehighestvolatilitytheHFTliquidityis about10%lessthanonanaverageday.

25

StructuralIssuesConcerningAlgorithmTradinginIndia

Graph2

HFTactivitymorethanaverage

HFTactivityLessthanaverage

Source:HighFrequencyTradingandItsImpactonMarketQualitybyJonathanBrogaard, Sept2010 OntheleastvolatiledaysHFTtakeabout7%lessliquiditythannormal,andonthemost volatiledaystheytakearound5%moreliquiditythannormal. Graph3

HFTactivitymorethanaverage

HFTactivityLessthanaverage

Source:HighFrequencyTradingandItsImpactonMarketQualitybyJonathanBrogaard, Sept2010 TheseresultsshowthatHFTactivitydoeschangewithvolatility,butnotprecipitously.In particularonthemostvolatiledays,HFTdonotpulloutofthemarket.Onthesedays,there doesseemtobea510%transferofHFTactivityfromsupplyingliquiditytodemanding liquidity.

26

StructuralIssuesConcerningAlgorithmTradinginIndia

9.4DoesHFTcontributetoPriceDiscoveryProcess?
Ofthe118stocks,68stocksshowashavinggreatercontributiontopricediscoveryprocess and28ofthosestocksHFTNonHFTcontributionisstatisticallysignificant.Inthe50stocks, wherenonHFTcontributionisgreaterthanthatofHFT,thedifferenceisstatistically significantfor7firms. OnaverageHFTcontribute86%,[(0.1950.105)/0.105*100],moretopricediscoverythando nonHFT.BasedontheseresultswecansafelysaythatHFTcontributetopricediscovery process. Table2:HFTnonHFTVarianceDecomposition

Source:HighFrequencyTradingandItsImpactonMarketQualitybyJonathanBrogaard, Sept2010 WhereHFT%shows Variance(EfficientPrice,HFTtradeprice)

27

StructuralIssuesConcerningAlgorithmTradinginIndia

NonHFT%shows Variance(EfficientPrice,NonHFTtradeprice) TStat tstatisticsfordifferencebetweenHFTandNonHFTcontributiontopricediscovery Conclusion:HFTcontributemoretopricediscoverythannonHFT.

9.5WhatroleHFTplaysinprovidingmarketliquidity?
ThissectionanalyzesHFTandthesupplyofliquidity.

a) DoHFTordersprovideinsidequotesmoreoftenthanNonHFTorders?
Insidequoteiseitherbestbuyquoteorbestsellquote Table3:TablereportsnumberofminutesduringwhichHFTsareatthebestbidoroffer. (ThetimeduringwhichbothHFTsandnonHFTsarebothatthebestquotesisalsoincluded inthevalue.)

Source:HighFrequencyTradingandItsImpactonMarketQualitybyJonathanBrogaard, Sept2010

28

StructuralIssuesConcerningAlgorithmTradinginIndia

PanelA allstocksatalltimes, PanelB thestocksthatareofferinglowerspreadsthanaveragealltime PanelC thestocksthatareofferinglowerspreadsthanaveragealltime IneachPanelthedataaredividedintothreegroups,with40firmseach,basedonfirmsize Viz.Small,MediumandLarge Sample=780=2(Providingeitherinsidebid/offerevery minute)*60(minutes/hour)*6.5(tradinghours)=NumberofminutesthataHFTcould potentiallybeprovidingatleastoneinsidequote PanelAshowsthatHFTfirmsfrequently(ataround65.3%(=509.3/780*100)ofthetimeof theday)providethebestbidorofferquotes.AsthefirmsizeincreasesHFTsaremore competitiveintheirquotes,matchingorbeatingnonHFTsquotesforasignificantportionof theday.AlsothereisnosignificantdifferencebetweenliquidityprovidedbyHFTfirmsin Highspread&Lowspreadstocks Conclusion:HFTfirmsprovidebestbid/bestoffermorefrequentlythannonHFTfirms.

b) DoesHFTprovidedepthtotheorderbook?
Inprevioussection,theanalysisonliquidityhasbeenbylookingatthebestinsidebidand asks.AnotherwayoflookingatHFTimpactonliquidityisbylookingatthedepthofthe booksuppliedbyHFT.PanelAandBshowsincreaseinimpactcostduetowithdrawalof eitherHFTordersorNonHFTorders. Firmsaredividedbasedontheirmarketcapitalizations.VerySmallincludesfirmsunder$ 400million,Smallarethosebetween$400millionand$1.5billion,Mediumarethose between$1.5billionand$3billion,andlargeareforfirmsvaluedatmorethan$3billion. Dollar Dollarincreaseinimpactcost Basis Percentincreaseinimpactcost PanelAshowstheresultsofremovingHFTfromthebook.Asthetradesizeincreases,the priceimpactincreasesacrossfirmsofallsizesandforalltentradesizeincreases.TheSmall categorytendstobemoreimpactedbythewithdrawalofHFTliquiditythanistheVery Smallcategory. Table4:

29

StructuralIssuesConcerningAlgorithmTradinginIndia

Source:HighFrequencyTradingandItsImpactonMarketQualitybyJonathanBrogaard, Sept2010 PanelBshowstheresultsofremovingnonHFTsfromthebook. AcrossallcategoriestheremovingofnonHFTshasamuchlargerimpactthandoesthe removalofHFTs.ThismeansthatalthoughHFTssupplyliquidityin41%ofalldollarstraded, theyprovideonlyafractionofthedepthcomparedtononHFTs. Table5:

Source:HighFrequencyTradingandItsImpactonMarketQualitybyJonathanBrogaard, Sept2010

Conclusion:Beyondsupplyingliquidityin51.4%ofalltrades,HFTsfrequentlysupplythe insidequotesthroughouttheday.Whileremovingeithertypeoftraderwouldresultinprice impact,butremovingnonHFTshasalargerimpact

30

StructuralIssuesConcerningAlgorithmTradinginIndia

9.6DoesHFTgenerateordampenvolatility?
Assumption:Priceswouldhaveachievedtheiractuallevelsevenintheabsenceofhigh frequencytradesbutwouldjumparoundmore. Ofthe120firms,72ofthemhaveahighervolatilitywhenHFTrinitiatedtradesare removed.Further,AsmallmajorityoffirmsexperienceslightlyhighervolatilitywithoutHFTr initiatedtrades.Howeverofthese72stocks,onlyoneisstatisticallysignificant.This indicatesvolatilitywouldhaveincreasedintheabsenceofHFT. Ofthe48stockswheretheremovalofHFTrinitiatedtradesreducesvolatility,suggestiveof HFTscausingvolatility,noneshowastatisticallysignificantdifferenceinvolatility. Table6:

31

StructuralIssuesConcerningAlgorithmTradinginIndia

Source:HighFrequencyTradingandItsImpactonMarketQualitybyJonathanBrogaard, Sept2010 Conclusion:TheoverallresultsshowthatwhenHFTrinitiatedtradesareremoved,volatility increasesandthisdifferenceisstatisticallysignificant.Admittedlytheresultsarenotstrong inonedirectionoranother;theyleaninfavourofHFTreducingvolatility.

32

StructuralIssuesConcerningAlgorithmTradinginIndia

Chapter10

PossibleManipulationsUsingHFT/AlgorithmicTrading
1) FrontRunning
Frontrunningisanillegalactivityinwhichatradertakesapositioninanequityin advanceofanactionwhichhe/sheknowshis/herbrokeragewilltakethatwillmove theequity'spriceinapredictablefashion. ApartfromFlashTrades(discussedbelow),whichsomebelieveispartofFront Running,thereisnooccurrenceoffrontrunningusingAlgorithmicTrading. FlashTradingServicesshowstockpricinginformationtosomeselectmember around500millisecondsbeforeroutingittopublicmarket.Ifsuchordersareoflarge quantity,fewmemberscantakeadvantageofitbyfrontrunning. In2008,SEBIallowedstockexchangestoofferDMAfacility,whichmanybelieve wouldhelpreduceinstancesoffrontrunning.DMAfacilityallowsInstitutionswith largeorderstoexecutethetradesdirectlyonstockexchangewithoutmanual interventionofTradingMember.Thispreventsleakageofinformationandcurbsthe practiseoffrontrunning. ThereisnoconclusiveevidencethatwoulddefinerelationshipbetweenFront runninganduseofAlgotrading.Andhence,useofAlgorithmicTrading(without usingDMA)willnothaveanysignificantimpactonIncreaseordecreaseofnumber ofinstancesoffrontrunning.ButtheuseofDMAwillmostcertainlyreducenumber ofinstancesoffrontrunning.

2) UnintendedandUncontrolledoperationsofalgorithms
Sometimealgorithmsoperateinaberrantmanner,duetoseveralfactorslike programmingdeficiencies,networkproblems,unexpectedmarketscenariosandall hostsofendogenousandexogenousfactors.Suchaberrantoperationsometimes maycreateissuessuchasNetworkJam,ServerJam,Queingoforders, server/networkbreakdown,submissionandcancellationoforderswhichtraders havenointentiontotrade,Cancelationrequestsforordersthatwerenever enteredetc. Thoughthereareseveralsystemchecksbyexchangesinplace,thereisapossibility ofalgoactinginaberrantmanner.

33

StructuralIssuesConcerningAlgorithmTradinginIndia

Case1:NYSEimposedpenaltyof$150,000onCreditSuisseSecurities (USA)LLCforfailingtoadequatelysupervisedevelopment,deployment andoperationofproprietaryalgorithm. Beginningatapproximately3:40p.m.,aCreditSuisseproprietary algorithmroutedhundredsofthousandsofcancel/replacerequeststothe NewYorkStockExchangeforordersthathadbeenpreviouslygenerated bythealgorithm,but,duetoanunforeseenprogrammingissue,were neversentbythealgorithm.Theunusuallylargeamountof cancel/replacemessagescontributedtotheoverqueuingofmessage trafficinallofthesecurities,approximately975intotal,tradedatfive postsontheNYSETradingFloor.Messages,includingneworders, modificationsoforders,andcancellationrequestswerefrozeninqueue andcouldnotbeimmediatelyprocessed.Thesefivepostscouldnotbe closedontime,ultimatelyclosingbetween4:10p.m.and4:27p.m.

3) Orderstuffing/Quotestuffing
Quotestuffingisessentiallyadenialofserviceattack,aimedattryingtoslowdown amarketinanenvironmentwheremillisecondsmatter. Withquotestuffing,highfrequencytradersenteranenormousnumberofbidsand offers,significantlyoutsidethecurrentbidofferspread,justtointroduceavast amountofnoiseintothequotefeed.Allthatnoisetakestime(maybejustafew extrananoseconds)forrivalHFTshopsandExchangesTradingSystemstoprocess, givingthequotestufferacrucialtimeadvantage. TheSECislookingintowhetherquotestuffingexists,andwhetheritsastrategythat anybodyhasactuallyusedtomakemoney.Theyarealsoinvestigatingwhether QuotestuffinghasanyinvolvementinFlashCrashthathappenedonMay6,2010. InIndia,myguessisquotestuffingmaynotexistsatthispointintime,givenvery lowparticipationbyHFTsbutagainwecannotruleoutchanceofhappeninginthe futurewithmoreandmorepenetrationbyHFTs.Quotestuffingisworldwide regardedasveryseriousactandshouldbepunishedseriouslyifattemptedbyany marketparticipants.

4) FlashTrades
Flashordersshowsstockpricinginformationforabriefperiodtoalimitedgroupof membertraderswhocanthendecidewhethertofillanorderbeforeitisroutedout tothepublicmarket. DirectEdge,anelectroniccommunicationsnetworkbackedbyGoldmanSachs, CitadelandKnightTrading,wasthefirsttooffersuchanordertypetoitsmembers in2006.NASDAQandBats(U.S.exchanges)createdtheirownflashmarketinearly

34

StructuralIssuesConcerningAlgorithmTradinginIndia

2009inresponsetotheDirectEdgemarket.Bothvoluntarilydiscontinuedthe practiceinAugust2009. DirectEdgebecameaU.S.exchangeinJuly2010andusedflashtradingtogreat successinsiphoningmarketvolumeawayfromNYSE,BatsandNasdaqandnowit claimsthirdlargestexchangeinUS.CurrentlySECisconsideringbanningFlash TradinginUSfollowingcomplaintbyEuronextoperatorofNYSEandGletcoOneof thetopmarketmakersinUS. Officially,sofar,nostockexchangeinIndiaisofferingthisservicebutintroductionof SmartOrderRoutingmayopenthegatesofopportunitiesforstockexchangesto attractinstitutionaltradersbyprovidingflashtradingfacility.Flashtradingservice givesunfairadvantagetoselectgroupofpeople/institutionsattheexpenseretail investors.Thisserviceisdetrimentaltothedevelopmentofsecuritiesmarketand SEBIshouldenactregulationstobanitsoperationinIndia.

5) Internalisation/CrossTrades
Internalisationreferstoabilityofbrokerstomatchcustomersawayfrompublic markets. InIndia,SEBIhasbannedinternalcrossingofordersandmandatedthatallorders enteredbyclientsmustbesubmittedtocentralordermatchingsystemofthe exchanges.Thispreventssettingupofstealthstockexchanges,darkpoolsorprivate tradingvenues,stealingliquidityawayfromthemarket. Inspiteofhavingsuchstrongregulations,Icannotruleoutthepossibilityofuseof algorithmstodevelopInternalCrossingsystems,whichIamsuremanybrokerage firmswouldbeetchingtostartinordertoreducetransactioncostfortheir proprietarytrades.HighcostoftransactioninIndiacouldactuallyworkasan impetusforuseofsuchsystems. Iseenorealdangerinallowinginternalisationofproprietarytradingasitavoids extracostssuchasSTTonintradaytransactionsforTM.However,internalmatching ofclientsordersormatchingordersacrossdifferentTradingMembershouldnever beallowedasitraisesseveralClearingandSettlementissuesandalsothereisreal riskofnewunregulatedtradingvenuessuchasdarkpoolbeingsetup.

35

StructuralIssuesConcerningAlgorithmTradinginIndia

6) Layering
LayeringisatechniqueofenteringLimitOrderswithlargequantityatdifferentprice levelonanysideoforderbookinordertocreateafalseappearanceofbuyorsell sidepressure. Person,whowantstobuyastock,entershiddenbuyorderatbestbidandenters numerousselllimitorderwithlargequantityatsuchapricelevelsthathavelow probabilityofgettingexecuted.Thiscreatesanimpressionthattherearemany sellersinthescriptandhenceitmaygodowninducingotherpeopletoselltheir stockstobuyer(manipulator) InIndia,Layeringmaynotbeusedthatextensivelymainlybecauseabsenceof DeepOrderBook.Only5bestbidsand5bestoffersareprovidedinIndiafora particularscriptasagainstallpossibleordersprovidedinUS.Thislimitstheabilityof DayTraderstotakeacallbasedondemandandsupplyofthestock.Layering requiresasmanylimitorderstobevisibletoeffectivelyimplementit. Havingsaidthis,IstillsuspectmanycasesofmanipulationusinglayeringinIndiain LowVolumeorilliquidscriptswithoutthehelpofHFT.Permittinguseofalgorithmic tradinginLowVolumestocksmayaggravatethisproblem.Soifpermitted,itshould beverifiedthatLayeringalgosarenotpermittedtouse. Case2:FINRAimposed$1millionpenaltyonTrilliumBrokerageServices,LLCfor usinganillicithighfrequencytradingstrategy Trillium,throughnineproprietarytraders,enterednumerouslayered,nonbonafide marketmovingorderstogeneratesellingorbuyinginterestinspecificstocks.By enteringthenonbonafideorders,ofteninsubstantialsizerelativetoastock's overalllegitimatependingordervolume,Trilliumtraderscreatedafalseappearance ofbuyorsellsidepressure. Thistradingstrategyinducedothermarketparticipantstoenterorderstoexecute againstlimitorderspreviouslyenteredbytheTrilliumtraders.Oncetheirorders werefilled,theTrilliumtraderswouldthenimmediatelycancelordersthathadonly beendesignedtocreatethefalseappearanceofmarketactivity.Asaresultofthis improperhighfrequencytradingstrategy,Trillium'stradersobtainedadvantageous pricesthatotherwisewouldnothavebeenavailabletothemon46,000occasions. Othermarketparticipantswereunawarethattheywereactingonthelayered, illegitimateordersenteredbyTrilliumtraders.

36

StructuralIssuesConcerningAlgorithmTradinginIndia

7) Darkpools
Darkpoolsarevenueswheretradingisdoneoffofanexchangetoobtainprice improvementandnotmovethepriceifalargeblockofstocksisbeingtraded.The executionsdoneindarkpoolsareonlyhiddenfromthepublicandotherbrokers beforeandduringtrading.Darkpoolshavebeentraditionallybeenusedby institutionaltraders,whousethemtoexecutelargestocktransactionswithout movingthemarketagainstthemselves.Butdarkpoolshaveevolvedtogiveretail investorsandsmallerordersacrackatpriceimprovementandbetterliquidity. Advantagesofdarkpoolsarelowertradingcost,bettertradepriceandmore liquidityasbrokerhasmultiplevenuestooperatein.TheriskofDarkpoolsisthat wheninformationisleakedoutofadarkpoolitcanleadtofrontrunningbybroker andexecutingtradeaheadofotherDarkpoolparticipants.Alsofewinvestorsmay notbeawareofpriceavailableforsamestockelsewhere. Currently,DarkPoolsarenotoperatinginIndia.ButBlockDealplatformwhichis almostsimilartoDarkpoolplatformisoperationalinIndiaandInfactauthorisedby SEBI.LargeInvestorswhowanttobuy/sellmorethan5lakhsharesormorethan5 croreworthofsharescanenterorderinBlockDealplatform.Ifthoseordersare allowedtobeenteredashiddenandalgotradersarealsoallowedtooperateinthat space,thenwecanhavevirtualDarkPoolrunningonthesameexchange.Toprevent thiseitherordersshouldnotbeallowedtobeenteredashiddeninBlockDeal Segmentoralgotradersshouldnotbepermittedtooperateinthatspace.

8) WashTrades
Anillegalstocktradingpracticewhereaninvestorsimultaneouslybuysandsells sharesinacompanythroughtwodifferentbrokersinordertoincreaseturnoverand inturncapturetheattentionofgeneralpublicenticingthemtotradethestock. TheresearchconductedbyASXsuggeststhatthereisbeenanincreaseinthe numberoftradesbeingcancelled,majorityofwhichwereWashTrades,from0.16% inJan2009to0.39%inAug2009.ThereasonbehindthisismanyAlgoTradingfirms employdifferenttradingStrategiesvizarbitragestrategies,marketmaking strategies,directionalstrategiessimultaneously.Thisincreaselikelihoodthatbids andoffersenteredfromthosealgoswillinadvertentlyexecuteagainsteachother.
9)

LiquidityDetection
Liquiditydetectionisanumbrellatermfortradingstrategiesthatinvolvesending smallorderstolookforwherelargeundisclosedordersmightberesting,onthe assumptionthat

37

StructuralIssuesConcerningAlgorithmTradinginIndia

Whenasmallorderisfilledquicklythereislikelytobealargeorderbehindit.Some ofthecommonliquiditydetectionstrategiesare: Pinging:sendingoutlargenumbersofsmallorderswiththeintentionofgettinga fillortogaininformationaboutelectroniclimitorderbooks; Sniper:anAlgorithmthattriestodetecthiddenliquiditybytradinginroundor oddlotsuntilitcompletesorreachesaninvestorslimitprice; Sniffing:Usedtosniffoutalgorithmictradingandthealgorithmsbeingusedby sendingasmallportionofanorderwaitingtoseeifsomeonecomesandgetsit. Sniffersattempttooutsmartmanybuysidealgorithmictechniqueslikeiceberging. AsurveywasconductedbyTABBgrouptofindwhetherliquiditydetectionisaform ofmanipulation.74%ofmarketparticipantssaidthattheydidnotregardliquidity detectionasaformofmanipulation,becausethetraderdoesnothavedirect knowledgeofanorder.

10)

ExchangeshelpingtheirbrokersbyprovidingSpecialDataFeeds

Specialdatafeedsaretypeofdatafeedswhichareeitherprovidedearlytosome marketparticipantscomparedtoothersorwhichcontainsspecialinformationwhich isnotavailabletootherpublic.ThereispossibilityofthismalpracticeandSEBIwould neverknowaboutthissincedatafeedsprovidedbyexchangeareonlyforcolocated serversandnotforpublicorevenSEBI.Suchkindofpracticesmaycausesystemic breakdownofthefinancialsystem. Example:IfIknowsomeinstitutionismovinglargeblockofXYZshare,anditis sellingtilllowestpriceof800Rs.Icanfindthisopportunitytoshortstock/futurein ordertocoverlater.

38

StructuralIssuesConcerningAlgorithmTradinginIndia

Chapter11

CurrentRiskManagementframeworkinIndia

PriceBandsinEquityandF&OSegment
InIndia,therearepricebandsof2%,5%,10%and20%onindividualsecurity. Nopricebandisapplicableonscripsonwhichderivativeproductsareavailable.Restall securitiesarecategorisedindifferentcategoriesbasedontheirdailyvolumeandimpact cost.2%,5%,10%and20%bandsareapplicableonthosecategories. Oncescriphitapricebandoneitherside,tradingishaltedforremainingday. InF&O,therearenopricebandsbuttopreventerroneoustradeentry.Operationalprice bandsaredefinedas10%forIndexfutures,20%forstockfuturesandDeltabasedvaluefor Stock&indexoptions.

IndexwiseCircuitFilter/CircuitBreakers
IndexwiseCircuitFilterisanautomatedtradinghaltmechanism,whichstopstradingonthe exchangeforspecifiedperiodoftime. Thereare10%,15%and20%circuitbreakersontwomainindicesNifty&Sensex. 10%movement 15%movement 20%movement 9:15am1.00pm 1.00pm2.30pm 1HrmarketHalt 1/2HrmarketHalt 9:15am 1.00pm 1.00pm 2.00pm 2HrmarketHalt 1HrmarketHalt TradingHaltedforrestoftheday 2.30pm3.30pm NoTradingHalt 2.00pm3.30pm TradingHalted

MarginingSysteminEquityandF&O
IndianmarketsaremuchbetterplacedvisvistheUSmarkets,thankstothepracticeof collectingmarginsonarealtimebasis.Ifamemberfirmexhaustsitsmarginwithan exchange,itisbarredfromtakingfreshpositions.IfclientortradingmemberorClearing membercrossesspecifiedlimits,tradingfacilityissuspendedforallclientstradingthrough TM(incaseTMviolateslimit)andalltradingmemberclearingthroughCM(incaseCM violateslimit). StocksarecategorizedinGroupI,II&IIIonthebasisofimpactcostandstockstradedinlast 6monthsforthepurposeofassigningdifferentmarginsondifferentcategorieshaving differentrisk. InEquity,InitialMargin=VaRmargin+Exposuremarginiscollectedonupfrontbasis, TypicallyVaRMarginisintherangeof7.515%andExposuremarginrangesbetween5

39

StructuralIssuesConcerningAlgorithmTradinginIndia

10%.Sotypically,totalinitialMarginVariesfrom12.525%.MTMmarginiscalculatedby markingeachtransactioninsecuritytotheclosingpriceofthesecurityattheendoftrading. Andcollectedbeforestartofnexttradingday InF&O,InitialMarginandExposuremarginarecollectedupfront.InitialMargin=TotalSPAN Margin(VaRmargin)+BuyPremium+AssignmentMargin.AndExposureMargin3%ofthe notionalvalueofafuturescontract(forfutures)&Thehigherof5%or1.5standard deviationofthenotionalvalueofgrossopenposition(foroptions).TotalInitialMarginis typically1520%ofTransactionValue.MTMmarginsarecollectedbeforestartofnext tradingday.

PositionLimits
AttheendofeachdaytheExchangedisseminatestheaggregateopeninterestacrossall Exchangesinthefuturesandoptionsonindividualscripsalongwiththemarketwide positionlimitforthatscripandtestswhethertheaggregateopeninterestforanyscrip exceeds95%ofthemarketwidepositionlimitforthatscrip.Ifyes,theExchangetakesnote ofopenpositionsofallclient/TMsasattheendofthatdayinthatscrip,andfromnextday onwardstheclient/TMsshouldtradeonlytodecreasetheirpositionsthroughoffsetting positionstillthenormaltradinginthescripisresumed. InIndia,wehaveTMwise&clientwisepositionlimits,whichpreventasingleTMorasingle Investorfromtakinghugepositioninasinglestock.Forclientsitis1%ofTotalFreefloat capitalisation(forcash)and5%ofMWPL(forF&O) NSEAlsomonitorFIIandMFpositionsforsectoralandcompanyspecificlimitssetbyRBI& SEBI. Therearestiffpenaltiesforviolationofanyoftheabovementionedlimitsandmargining requirements.

PretradeControlinF&Osegment
QuantityFreeze:AnyordercomingtotheexchangefortradeinNIFTY,S&P500, BANKNIFTY,CNXITorMINIFTYfuture/optionwithquantitymorethan15,000willbeFreezed untilbrokerconfirmitasagenuineorder.Fortradeinfuture/optiononindividualstocks, freezequantityisasdecidedbyexchangefromtimetotime. PriceFreeze:PricefreezeisoperationalonlyintheF&Osegment.Operationalpricebands aredefinedas10%forIndexfutures,20%forstockfuturesandDeltabasedvalueforStock &indexoptions.

40

StructuralIssuesConcerningAlgorithmTradinginIndia

PretradeControlinEquitysegment
Apartfrompricebandstherearenopretradecontrolsinequitysegmenttocheck erroneoustrades. InUSerroneoustradesthatareexecutedwayoutsidelasttradedpricemaybecontested andcancelledaftercomplainingtotherespectivestockexchange.Ihavegiventwocases wheretradeswerecancelled. Case3:NYSEEuronextcancelledalltradesinthe$74.8billionSPDRS&P500ETFTrust thatoccurredatalmost10percentbelowthesecuritysopeningprice,accordingtoan emailsentbytheexchange. ThetradesoccurredontheNYSEsArcaplatformat4:15p.m.NewYorktimeon 18/10/2010andpricedtheexchangetradedfundthattrackstheStandard&Poors500 Indexat$106.46comparedwithitsopeningpriceof$117.74.TheETFplunged9.6 percentovereightsecondsas7.2millionsharestradedonNYSEArca,accordingtodata compiledbyBloomberg.TheS&P500rose0.7percenttocloseat1,184.71today.The glitchoccurredaftertheclosingauctionwasdelayedduetoasoftwareupgradeatthe exchange.Officialclosingpriceineachsecurityisdecidedinanauctionprocess conductedaftercloseofthemarket. Case4:Nasdaq&NYSEcancelledalltradesin296securitiesthatwereexecutedat greaterorlessthan60%oftheirpricesat2:40p.m.ETonMay6,2010 PersonallyIdontthinkanytradeshouldbecancelledevenifitisthepunchingerror. Howeveritisalsotruethat,unintendederroneoustradesshouldnotbepunishedsoheavily thatoperatorsareafraidtoenterorders.Onewaytocounterthiswouldbeintroducing tighterpretradecontrol.Ihaveexplainedintroductionof5%dynamicpricebandsaspre tradecontrolinrecommendationsectionofthisreport.

41

StructuralIssuesConcerningAlgorithmTradinginIndia

Whyisitimportanttoimplementsuchstringentoperationaldynamicprice bands?
ThereweretwocasesinthehistoryofIndianSecuritiesMarketthatcausedhugelosstothe brokingfirmsandtheirclientsduetopunchingerrorsbyoperators. Bothofthesecasescouldhavebeenavoided,hadtherebeen5%dynamicpriceband operational.Havingnarrowdynamicpricebandof5%ensuresthatasingletradecantmake anystockfallbymorethan5%. Case5:FirstcasewasofpunchingerrorinTulipITservicesonJan5,2006.TulipIT Servicessawtwounusualtradeshappeningonitslistingday.Onedealsaw4,04,800 sharessoldat25paiseagainsttheaveragepriceofRs185attheBombayStock Exchange,whiletheothersaw5,95,575sharessoldatRs100each. Thosedealscaused theselleralossofRs12.69croreinjust38seconds.Thefirsttransactionledtoanet lossofRs7.40crorewhilethesecondtransactioncostRs5.06crore. Case6:ThesecondcaseisfreaktradeinRelianceonJune1,2010.Amarketorderto sell1.6lakhshareswasenteredinReliancebyadealeratabrokermemberfirmwhich gotmatchedwithlimitpurchaseordersinthetradingsystem.Withinfewsecondsof enteringorder,ReliancestockplungedtoRs840.55frompretradepriceofRs1,010 causingscripttofallby19.56%.Manyanalystsbelievethistrademighthavecost brokingfirmamorethanRs1crore.Apartfromlosstothebroker,suchtradeshurt investorsconfidencewhichtakestimetorecoup.

42

StructuralIssuesConcerningAlgorithmTradinginIndia

Chapter12

Algoapproval,registrationandriskmanagementframework

Algorithmapprovalandregistration
BecauseSEBIhasnotissuedaspecificguidelinesregardingalgorithmtrading,thereis divergenceinoperationoftwomajorIndianexchanges.NSErequiresallalgorithmstobe approvedandallentitiesusingalgorithmstoberegistered.WhereasBSErequiresmere registrationtostartAlgoTrading.BSEdoesnotrequireitsmembertodisclosetheir proprietarytradingalgorithmstoit. ThoughBSEisfollowinganinternationalpracticeofnoninterventionbytheexchangesin useofalgorithms,itisquiteadangerouspracticetostartwithinanmarketwhichisnotyet fullydevelopedandanyeventsimilartoFlashCrashinUSmaywhiskawayretailinvestors fromsecuritiesmarketpermanently.AmericanandEuropeanexchangeshavedeveloped stateofarttechnologytopreventanymajorbreakdownofthesystemandinspiteofthese measuresbreakdownhappens.AndATisatoolifnotproperlyimplementedandmonitored cancausehavoc.Soitisbettertobeoncautioussidethantorepentlater. SoIsuggestthatSEBIshouldissueguidelinestoexchangesrequiringthemtoapprove algorithmspriortogivingthempermissiontooperate.Approvalwillbeonthebasisof implementationofriskcontrolmeasures.

ProcedureforgrantingpermissiontoTradingMemberforAlgorithmTrading atNSE
a) NSE,onMay17,2011,issuedacirculartosmoothentheprocessandfacilitateearly approvalofdecisionsupporttools/algorithmsfortradingthroughnonneatfront end.Itcategorisedapplicationforseekingapprovalinalgotradinginthreemain categoriesviz.ApprovedAlgorithmsthroughCTCL,NonApprovedAlgorithms throughCTCLandDMAAlgorithms.ThiscategorisationhelpsNSEtoscrutinizenew algosmorefortheirpretradeandposttraderiskcontrolsthanalreadyapproved algos. b) OnfulfilmentofconditionsassatisfactoryandmeetingSEBIandExchangesminimum requirements,theexchangewillidentifythevendors/membersalgorithmas Approvedalgorithm. c) Totalproceduretakesaround30daystogetalgoapprovedfromdayofapplication. (FordetailsrefertoannexureII) BSEdoesnotrequireAlgorithmapprovaltostartalgorithmictrading

43

StructuralIssuesConcerningAlgorithmTradinginIndia

RiskManagementFrameworkatthetimeofalgorithmapproval
CurrentlyNSEhasfairlystrongcheckstoapprovealgorithms/decisionsupporttools;the onlythingweshouldbeconcernedaboutisitsimplementation,fairnessinimplementation andtransparency. Thealgoapprovalprocessincludesfollowingchecksatindividualandclientlevel: Table7 S No. Requirement Whether complied (Yes/No) YES/NO YES/NO YES/NO YES/NO YES/NO

1 Doessoftwareperformfollowingchecksbeforeroutingorderstothe exchangeforexecution? 2 3 4 5 IndividualOrderLevel: QuantityLimitscheck PriceRangechecks Tradepriceprotectionchecks Ordervaluechecks ClientLevel: Netpositionv/savailablemargins RBIviolationchecksforFIIrestrictedstocks MWPLviolationchecks Positionlimitchecks Tradinglimitchecks Exposurelimitchecksatindividualclientlevelandatoveralllevelforall clients DoesthesystemhaveProvisionforgeneratingandmaintainingcomplete AuditTrail. CantheordersgeneratedbyAlgorithmictradingproductsbeidentifiedas AlgorithmorderswhilereleasingtotheExchange AllordersgeneratedbyAlgorithmictradingproductsareofferedtothe marketformatchingdirectlyandnocrosstradesaregenerated? Doesthesystemhavesufficientsecurityfeaturesincludingpassword protectionfortheuserID,automaticexpiryofpasswordsastheendofa reasonabledurationandreinitialisationofaccessonenteringfresh passwords?

44

StructuralIssuesConcerningAlgorithmTradinginIndia

Chapter13

Concerns
1. AreretailinvestorsareatdisadvantageduetouseofATandHFT mostlybyinstitutionalinvestors?
Certainlynot.Firstreasonistypicallyinvestorsinvestmenthorizonislongterm,so millisecondsadvantage,whichcomesatexorbitantcost,mayhardlymattertothem. Secondeveniftheyhavemillisecondoftimeadvantage,theyarenot100%sure aboutfuturepricemovement.Whattheyknowisprobabilityoffutureprice movementwhichanyretailinvestorhasanoptiontocalculateonhisownbyusing freelyavailabledataservices.ThirdHFTdoesnotcompetewithlongterminvestors, theycompetewithotherHFTS.AsmoreandmoreHFTcomeinthemarket,they havelessspreadstoenjoy;theywillhaveveryshortlivedarbitrageopportunities.So theywillcompetemoreonspeed.Ifnowtheyexecuteordersinmilliseconds,in futuretheywilldeveloptechnologythatwillexecutetradesinmicroseconds.Apart fromthisHFTsreducesspreadandimprovesliquiditybringpricesinallmarketon parandcontributetopricediscoveryprocess.Inabsenceofmanipulative/predatory strategiesbyHFT,retailinvestorwillnotbeatdisadvantage.

2. IsallowinguseofDMAfacility,COLOfacilityandSORfacilitygivean unfairadvantagetosomeofthemarketparticipants?
InIndia,nonintermediaryDMAfacilityisnotallowed,whichmeansalltheordersof clientshavetogothroughbrokerandwillbetaggedasthatbrokersorder.Sothere islessscopeofmanipulationasordergenerallygoesthroughbrokersrisk managementsystem.Brokerisnotabletostopormodifytheordermanuallybutits systemcanrejectthebadordersupfront.ThisleavesonlyonewaytomisuseDMA, i.ewhenbrokeritselfrelaxeshisautomatedriskcontrolsforhisclient. Colocationisserviceofferedbyexchangestotheirtradingmemberstosetuptheir serverscloseinproximityofexchangesordermatchingsysteminordertoreduce networklatencyandaccesslatency.Theserviceoffersvariedadvantageincluding lowturnaroundtime,whichisverycrucialforHFT,alternatesetupincaseof networkorsystembreakdown.Colocationfacilityoffersusually12milliseconds advantagetotheirTMs,whichreducestradinglatency.Soasstatedin1stconcern,1 2millisecondadvantagemaynotmattertolongterminvestorsandthattooat exorbitantcost.Theproblemwiththisservicestartswhenexchangescolludewith brokerstoincreasetheirbusinessbyadoptingunethicalandunfairpracticessuchas providingspecialdatafeeds,givingpriorityinordersubmissiontocollocatedusers, etc.

45

StructuralIssuesConcerningAlgorithmTradinginIndia

SORfacilityisalgorithmthatautomaticallyroutstheordertotheexchangewhich offersbestbid/bestofferforinvestor.Thisservicehasbeenaroundforquitesome timenowinUS&Europeanditcontributessignificantlytopricediscoveryprocess. SORisconcernwhenitisusedforPinging/Sniffingoutlargeorderatsomeother exchangeoratsomeothersegment(blockdealsegment)andthentradeaheadof institutionalorder.TheotherconcernofSORwouldbewhenexchangesputsome restrictiononitsusetopreventflowofitsbusinesstosomeotherexchange.Overall advantagesoverweighdisadvantages.

3. DoesAlgorithmTradingimproveMarketQuality?

AsdiscussedinthetopicImpactofHFTonmarketquality,HFTinUshashelped reducespreadsandimproveliquidity.AlsoHFThasincreaseddepthoftheorder books.Howevertheymightnothaveconsiderableimpactonchangeinthevolatility. TheeffectofHFTonvolatilitywasinconclusivebutstillwasleaninginthefavourof HFT.Manyresearchersandexchangeshavecorroboratedthepointthatthereisno significantcorrelationbetweenparticipationofHFTsandvolatility.Andeventsuch asFlashcrashcouldhaveoccurredevenintheabsenceofHFTs. SinceHFTispartofAlgoTradingandpresentsmoresevereriskthannonHigh Frequencyalgotrading,wecansafelyextentconclusionsofabovestudytoATas well. HereweshouldmakeanoteofonepointthatsignificantnumberofHFTsinUSare marketmakershaveemployedrebatetradingstrategiesduetofavourablemaker takerstructure.SoassumingthatliquiditywillimprovealsoinIndiabecauseof introductionofATwouldbeincorrect.

4. IsHFTandATinthelongterminterestofthemarketdevelopment?

Theanswerdependsuponfairnessandethicsofthemarketparticipantsincluding exchanges.Fewmaytwistandbendtherulestosuitthem.Butthisshouldnotstop usfrommovingaheadwithtime.Marketdevelopmentsshouldnotbestoppedjust becausewefearsomemaymisusethesystem.Soitisalwaysadvisabletostartwith cautiousapproachbyslightlyregulatingmarketwithalongtermviewofrelaxingthe regulations.SEBIcannotandshouldnotregulateeachandeverything.Itisbeyond itsscope.ButatanypointintimemessageshouldbeloudandclearWhateveryou do,Wearewatchingyou. TherearequiteafewadvantagesofAT,whichincludesincreaseinliquidity, decreaseinthespreads,betterpricediscovery,moreparticipationbyOversees

46

StructuralIssuesConcerningAlgorithmTradinginIndia

Institutionsanditcanbeusedasriskmanagementtool.Tosumitup,IcansayAT andHFTareinthelongterminterestofthemarket.Itsadvantagesoutweigh disadvantagesandIbelieveitistheneedofthetimeandifwedontchangenow timewillchangeus.

5. IsinfrastructureinIndiaadequateenoughtohandleATandHFT?
Ifpublicallyavailableinformationistobebelieved,wehaveadequatetechnological infrastructuretoallowAToperationsbutcertainlyitsnotoneofthebest.Wefall shortonstandardssetbyNYSE,NASDAQandfewotherEuropeanexchanges. Forexample,Latency(responsetime)atNYSEandNASDAQare25microsecondsand 143microsecondsrespectivelycomparedto5,000microsecondsand10,000 microsecondsatBSEandNSErespectively.SinceIndiaisthe4thlargestcountryinthe worldintermsofnumberoftrades,Indianservershavebeenalwaysoverburdened andintroductionofATandHFTwillmakethingworseasitwillputanadditionalload onexistingoverburdenedsystem.

6. Dowehavesufficientlyrobustrealtimeriskmanagementsystem?
FornonalgorithmictradingandnonHFT,wehavesufficientstrongriskmanagement systemwhichincludespretradeordercontrols,priceandquantityfreeze,security wiseandmarketwisecircuitbreakers,realtimemargincollectionsystem,and barringclient/TM/CMfromtakinganypositioniftheirlimitsareexhausted. Inalgorithmictradingthistaskmaygetlittlecomplicatedowingtotremendous speedofexecutionsandDMAfacilitywhichmaybypassTMsriskmanagement framework.Tobedoublysure,itisworthwhiletosurveyandauditexchangesRisk Managementsystemsfromalgotradingperspective.Itisalsorequiredthat exchangesshouldprovidetheirRiskcontrolchecklisttoSEBIsothatSEBIcould decideminimumriskcontrolchecksforalltheexchangesbeforegivingapprovalto algorithms.

7. ArethereanysystemicriskarisingfromAlgoTrading?

Thereispossibilityofsystemicrisks,ifexchangesarehandinglovewiththeirTrading members.Ifexchangesprovidespecialfeedstobrokerswhohavecolocatedtheir serversatexchanges,thenitgivesunfairadvantagetosomemarketparticipants.

47

StructuralIssuesConcerningAlgorithmTradinginIndia

ApartfromthiswidespreaduseofanyofthefraudulentpracticessuchasFront Running,Orderstuffing/Quotestuffing,FlashTrades,Internalisation/CrossTrades, Layering,DarkpoolsandWashTradesmaycausesystemicfailureoffinancialsystem inIndia.

48
StructuralIssuesConcerningAlgorithmTradinginIndia

Chapter14

Recommendations
1. SEBIshoulddevelopordersegregationplatform/orupgradean IntegratedMarketSurveillanceSystem"(IMSS) tosegregateATorder, HFTorder,COLOorder,DMAorderandSORorderindividuallyorin combination.
Exchangesaresayingthattheyhaveassigneduniquedigitcodeforeach15digit orderandhenceorderscaneasilybesegregated.Ifso,SEBIshouldbeintheposition tosegregateorderswithanycombinationofordersandentities/persons. ForexampleOnaparticularday,SEBIshouldbeinthepositiontoidentifyNumber oforderscancelledbyaXYZHFTfirm,whichwereenteredthroughDMAfacility

IfFlashCrashlikeeventhappensinIndia,Howwillweinvestigatethematterifwe donthavedata?HowwillweknowwhichalgotraderssoldXnumberofsharesand atwhattime?Howwillpossiblyweknowaboutongoingmanipulationsor malpracticesifwecantsegregatetradesbasedontheirtypeoforigin?Howwillwe beabletopinpointaspecifictypeoftradingthatcausedhavoc?Andhowwillwebe abletodevelopguidelinesifregulatoritselfiscluelessaboutmarketconditions?

2. Introductionandstrengtheningofthreetierpretradecontrols

Pretradecontrolsareproactiveinnatureratherthanreactive.Pretradecontrolshelpavoid entryoferroneoustradeandviolationoflimitsbydifferentparticipants.Presubmission orderlevelfiltersandPresubmissionaccountlevelfiltersareemployedatbrokersend. Thesemeasureshelpexchangestoreducelatencyorroundtripexecutiontimeasmoreand moreorderswillberejectedatbrokersendonly. Presubmissionorderlevelfilters:Filtersthatassessanorderasadiscreteunit ratherthanaspartofanaccount.Filtersaredesignedtopreventerrorsinprice, valueanddirectcompliancesuchasshortselling. Presubmissionaccountlevelfilters:Filtersthatassessanorderinthecontextof accountlevelexposure.Suchfiltersdependonaconsolidateaccountposition irrespectiveorexecutionvenue. Pre-execution Exchange level filters: Anyordersubmittedtoexchangeshouldbe authenticatedforquantity,price,clientwiseandTMwiselimitsandanyother relevantcriteria.Italsochecksforpricebandsandmarketwidecircuitfilters.

49

StructuralIssuesConcerningAlgorithmTradinginIndia

3. Introductionof5%dynamicpricebandsonallsecurities(Oneofthe Presubmissionorderlevelfilter)
Introducingpricefilterstopreventsystemsfromenteringerroneoustrades. Dynamicpricebandshouldbedefinedas5%ofthelasttradedpriceinall scriptslistedatNSE&BSE. Thispricebandwillchangewitheverynewtradehappening.Ifchangingprice bandswitheverynewtradeisnotfeasible,wecanimplementiton5seconds delayedbasis.Sopricebandwillchangeevery5seconds. Justforexample:IfaXYZstockiscurrentlytradingatRs100,Anyalgo tradingsystemwillnotbeabletoenterbuyorderatmorethanRs105and sellorderatlessthan95.Howeverbuyordershallbeallowedtoenterbelow 95andsellordershallbeallowedtoenterabove105.Placingabuyorder belowcurrentmarketprice&placingasellorderaboveCMPwillimprove liquidityasthiswouldincreasenumberofpassiveorderssittingintheorder book. ThisruleshouldonlybemadeapplicableforLimitOrdersandMarket Ordersandshouldnotbemadeapplicabletostoplossorders.InstopLoss ordertradercanplacetriggerpriceandlimitpriceatanypricehedesires. Thispresubmissionorderlevelcheckwillbevalidatedatbrokers(Trading Members)endandtheorderwillberejectedoutrightandwillnotbe submittedtoexchange. Thisruleshouldalsobeextendedtoallmanuallyenteredaswellasnon algorithmicorder. ThisruleshouldalsobemadeapplicableonthefirsttradingdayoftheIPO listing. Thisruleensuresthatanysingletradecantmakeanystockfallbymorethan 5%irrespectiveofquantityandpriceentered.

4. Mechanismtoflushoutallthependingorders
OnMay6,2010,DowJonesIndextumbledmorethan6%injust4minutesdueto relentlesssellingbyHFTsinEMini500contracts.Inordertopreventacascadeof furtherpricedeclines,CMEtriggeredSTOPLOGICFUNCTINALITYwhichhalted tradinginEminifuturecontractsfor5seconds. Inthatshortperiodoftime,sellsidepressureintheEMiniwaspartlyalleviated andbuysideinterestincreased.Whentradingresumedat2:45:33p.m.,prices stabilizedandshortlythereafter,theEMinibegantorecover.SaidSECCFTCjoint report.

50

StructuralIssuesConcerningAlgorithmTradinginIndia

STOPLOGICFUNCTINALITYisbasicallydisastermanagementtoolandevery exchangeshouldhaveitinitarsenal.IexpectNSE/BSEalreadyhaveitandifnotitis thetimetodevelopone.

5. IntroductionofMakerTakerStructurefeetoimproveliquidity
MakerTakerstructureinvolvesgivingcredittobuyerorsellerforprovidingliquidity toothermarketparticipantsandchargingfeefortakingliquidity.Liquidityis providedbyenteringpassivelimitorderandliquidityistakenwhenyouhitthebidor taketheofferbyenteringmarketorder. Anexampleofanewtradingstrategythathasemergedoverseasandattractedthe labelpredatoryisaFormofarbitragedesignedtoensurethattheHFTreceivesa rebateforprovidingliquiditytoaretailorinstitutionalinvestor.Thisisalsocalled RebateTrading. InIndia,BSEhasintroducedMakerTakerstructureinitsF&Osegmentbutdidnot haveanysignificantimpactinitsmarketshareasitalreadyhadverylittlemarket share. Transactioncharges/rebateforevery1LakhsharesenteredasPassive/Activeorders areasfollows Passive ActiveOrders Orders Futures(Stock&Index) Credit1Rs Charge1.5Rs Options(Stock&Index) Credit15Rs Charge20Rs PassiveOrders:Passiveordersaredefinedastheordersthatalreadyexistinthe orderbookatthetimeofmatching(ordertakingplace). ActiveOrders:Activeordersaredefinedastheordersthatarematchedagainstthe ordersalreadyexistingintheorderbookatthetimeofmatching(tradetaking place).

6. AskingexchangestoprepareareportonStructuralissuesand systemicrisksarisingduetoAlgorithmTrading
Itisaworldwidepracticetoupdatetheregulatoraboutoperationalproblems, potentialthreatsandlatesthappeningontechnologyanddevelopmentfront.As exchangesareindirectcontactwithmarketparticipantsandalsoknowminuscule detailsabouttheirordermatchingplatform,theywouldbemoreequippedandin betterpositiontohelpSEBIachieveitsgoalsofmarketdevelopmentandinvestor protection.

51

StructuralIssuesConcerningAlgorithmTradinginIndia

LondonstockExchange,AustralianSecuritiesExchangeandmanyotherstock exchangeshavepreparedthereportonabovementionedtopicandareavailableto thepublic.

7. AuditandSurveyofExchangesriskmanagementsystemsfor algorithmictradingbyaspecialteamandsubmitthereporttothe highermanagementonthesame

Onpaper,NSEandBSEhavesoundRiskmanagementSystem(RMS)butweneedto ensurethata)systemisfaultless,b)Capabletomonitorandcontrolalltradesc)is notorleastsusceptibletoattacksontheirservers/ordermatchingplatformsd)is notpronetofrequentbreakdownande)TheyhavealternateRMSincaseof breakdown. Itisessentialtoassessalltheriskcontrolparametersandpresentriskmanagement systembeforeproposinganypolicychanges.Indianstockexchangesarenotthat keenonprovidingextrainformation.IfSEBI,beingaregulator,doesntknowabout theircurrentRMSthenpossiblynobodywouldknowaboutituntilcatastrophe occurs.SoIhope,SEBIwouldsendaspecialteamtoinvestigateallissuesfromAT perspectiveandgivenachance,Iwouldliketobeapartofthatteam.

8. DefiningsomeofthepracticessuchasFrontRunning,Orderstuffing /Quotestuffing,FlashTrades,Internalisation/CrossTrades,Layering, Darkpools,WashTradesandspecialdatafeedsasfraudulentand unfairtradingpractices,banningitsuseanddefiningpunitiveactionif violated.

9. DataReportingrequirement
NSEandBSEtoprepareMonthlyReportonAlgoTradinginformatsspecifiedin Table8andshouldprovideittoSEBIasandwhenasked.Theyshouldprepareand analyseATfirmwisereportregularly,evaluaterisksinvolvedandbriefSEBIincase thereisanyconcern.Sinceinformationintable8iscompanyspecific,wecannot publishitinourMonthlyBulletinandannualReports.Butthisdataisstillneededto assesscompanyspecificrisks,toinvestigatescamsandtodeveloppolicies. FortheinformationofInvestors,exchangesneedtoprovidedatatoSEBIinformat specifiedinTable9onmonthlybasis.ThisdatawillhelpSEBIensurethatthe requiredinfrastructureforAlgoTradingisinplaceandalsoitwillprovideInvestors anoptiontoexitthemarket,iftheyfeelthereisheightenedriskduetohighamount ofAlgoTradingActivity.

52

StructuralIssuesConcerningAlgorithmTradinginIndia

Table8 Nameof theFirm 1 Average Size(Qty) ofthe order Typeoffirm Dateof Numberof (BuySide/ approving algos Sellside) 1stalgo approved 2 TotalTurnover (Monthly) Through Passive Orders 11 Optedfor DMA facility? 12 3 4 Orders entered 5 Orders Cancelled 6 Orders Traded 7

Average Valueofthe Through Order Active Orders 8 9 10

OptedForCo Location?

13

Table9 Capacity(Orderspersecond) Latency(RoundtripExecutiontime) Latency(Datafeeds/Infodisseminationtime) AverageDailyOrders OrderFlow(throughDMAfacility) Orderflow(throughAlgoTrading) AlgoTradingvolumeasPercentageofTotal Volume OrderFlow(throughCoLocationfacility NumberofordersdivertedtootherStock ExchangesthroughSOR NumberofAlgosregistered/approved NumberofBuysideFirmsregisteredforAT NumberofSellsideFirmsregisteredforAT MaxNumberofOrderspersecondperAlgo MaxNumberofOrderspersecondperAT firm AverageOrderToTradeRatio AverageOrderCancellationRate

BSE

NSE

53

StructuralIssuesConcerningAlgorithmTradinginIndia

Chapter15

Conclusion

AlgorithmTradinghasbeenacceptedinternationallydueto:reducedmarketimpact; increasedtradingefficiency;betteralignmentbetweenstrategyandexecutionandlower cost.ATisalsoseentodrivedowntradingcosts.ItisestimatedthattheuseofATreduces transactioncostsbyalmost75%.Anothersignificantadvantageistheanonymityof algorithmswhichhelpspreventinformationleakage.Further,algorithmicstrategiessatisfy regulatorycompliance.Italsoenhancestheefficiencyofthefinancialmarketfor arbitrageursandeliminatesfrontrunning. WhilenegativesofATMayIncludeunderminedmarketstabilityduetofrequenttradingand systemoverloadissues,whichneedtoberesolvedonurgentbasis. HFTwhichispartofAlgorithmTrading,asobservedintheinternationalmarkets,tendstoa) IncreaseLiquidityb)Reducespreadsc)Contributetopricediscoveryprocessandd)haveno impactonvolatility. Thelargerconcernabouthighfrequencytrading(HFT)isthatitcoulddestabilizemarkets, especiallyaftertheflashcrashintheUSon6May2010.Manyanalystsbelievethatsucha situationcanariseeveninaworldwithoutHFTandweshouldnotpunishHFTforsuchone offinstanceofheighteneduncertainty.Aflashcrashsituationariseswhenalargeorderis placedintoathinorderbookinadvertently. Fewmaytwistandbendtherulestosuitthem.Butthisshouldnotstopusfrommoving aheadwithtime.Marketdevelopmentsshouldnotbestoppedjustbecausewefearsome maymisusethesystem.Soitisalwaysadvisabletostartwithcautiousapproachbyslightly regulatingmarketwithalongtermviewofrelaxingtheregulations.SEBIcannotandshould notregulateeachandeverything.Itisbeyonditsscope.Butatanypointintimemessage shouldbeloudandclearWhateveryoudo,Wearewatchingyou.

54

StructuralIssuesConcerningAlgorithmTradinginIndia

AnnexureI RiskmanagementcriteriasforapprovalofDMAfacilityatNSE Whether complied (Yes / No)

S. No. 1 2 3

Requirement Access for order routing is permitted only through the use of client specific User ids The access is permitted only through the use of client specified password (private code) The system provides for automatic expiry of passwords at the end of a reasonable duration and reinitialisation of access on entering fresh passwords. All transaction logs are maintained with proper audit facilities alongwith unique numbering for orders/trades (a write-up in this regard shall be enclosed) Logic/priorities similar to those used by the Exchange is used to treat client orders. Provision for maintenance of all activities / alerts log with audit trail available Authorized user and client details are part of the order details received and authenticated at the DMA server of the trading member. The members Direct Market Access (DMA) server internally generates unique numbering for all client orders/ trades. Adequate provisions for maintaining back up systems and data storage capacity are available. An alternative means of communication is arranged for in case of link failure ( A write up shall be provided in this regard) The software provides for appropriate authority levels to ensure that the limits can be setup only by the persons authorized by the risk / compliance manager. The software provides for routing of orders through electronic / automated risk management systems of the broker to carry out appropriate validations of all risk parameters including Trading Limits, Position Limits and Exposure taken by client based on risk assessment, credit quality and available margins of the client. The software also provides for appropriate validations for Quantity Limits, Price Range Checks, Order Value, and Credit Checks before the orders are released to the Exchange. The software provides for Net position that can be outstanding so as to fully cover the risk emanating from the trades with the available margins of the specific client. The software provides for appropriate limits for securities which are subject to FII limits as specified by RBI. The software provides Direct Access Market (DMA) facility

5 6 7

9 10

11

12

13

14 15

55

StructuralIssuesConcerningAlgorithmTradinginIndia

16

17

to clients only to execute his own trades and shall not use it for transactions on behalf of any other persons/ entity. All orders entered into the Direct Market Access (DMA) system are offered to the market for matching and no cross trades are generated. The software has the facility of providing the reports on margin requirements, payment and delivery obligations etc. to the clients through the system.

56

StructuralIssuesConcerningAlgorithmTradinginIndia

AnnexureII ProcedureforgrantingpermissiontoTradingMemberforAlgorithmTrading 1. Procedureformakinganapplication Tradingmemberdesirousofseekingapprovalfortradingusingalgorithmsshouldmake anapplicationinformatasspecified.NSEcategorizedalgorithmapplicationsfor approvalinthreecategories,whichareasfollows d) ApprovedAlgorithmsthroughCTCL TradingMember,desirousofusingalgorithmswhicharealreadyapprovedbyNSEto thevendors,canmakeanapplicationinthiscategory.Memberneednotgive demonstrationoftheAutomatedRiskmanagementfeaturesasthatpartisalready donebytheVendor. e) NonApprovedAlgorithmsthroughCTCL TradingMember,whointendstousealgorithmswhicharenotapprovedbyNSE, mustmakeanapplicationinthiscategory.TradingMembermustgive demonstrationoftheAutomatedRiskmanagementfeatures.Andsubsequently, membershallberequiredtoinformNSEincaseofanychangetoapproved algorithms. f) DMAAlgorithmsTradingMember/InstitutionalClientsAlgorithms Irrespectiveofwhetheralgorithmisapprovedornot,TradingMembersor InstitutionalclientswhointendtousealgorithmthroughDirectMarketAccess facility,mustmakeanapplicationinthiscategory. 2. Procedureforalgorithmapproval OnfulfillmentofconditionsassatisfactoryandmeetingSEBIandExchangesminimum requirements,theexchangemayidentifythevendors/membersalgorithmas Approvedalgorithm. 3. AdditionalCategoryofNonNEATuserswillbeintroduced ExchangeisplanningtoprovideseparateNonNEATuseridstothememberswhoare usingNonNEATfrontendfortrading.Suchmemberswillnotbeallowedtologin/trade throughNEATbutwilluseitonlyforriskcontrolandtradevalidation. ThisisverypositivesteptakenbyexchangetohelpmemberswhoarenotusingNEAT fortradingtomanagerisks. 4. RiskManagersforAlgorithmTerminal Eachriskmanagersshallbeallowedtomanagemultiplealgorithmsundersinglerisk managementsystem.Henceitwouldbeconsideredassingledealerforthecompliance purpose.

57

StructuralIssuesConcerningAlgorithmTradinginIndia