The Science of Algorithmic Trading and Portfolio Management

474 Pages·2013·5.35 MB·English

by  Robert Kissell (Auth.)

The Science of Algorithmic
Trading and Portfolio
Management
The Science of Algorithmic
Trading and Portfolio
Management
Robert Kissell
Ph.D
AMSTERDAM(cid:129)BOSTON(cid:129)HEIDELBERG(cid:129)LONDON
NEWYORK(cid:129)OXFORD(cid:129)PARIS(cid:129)SANDIEGO
SANFRANCISCO(cid:129)SINGAPORE(cid:129)SYDNEY(cid:129)TOKYO
AcademicPressisanimprintofElsevier
AcademicPressisanimprintofElsevier
525BStreet,Suite1800,SanDiego,CA92101(cid:1)4495,USA
TheBoulevard,LangfordLane,Kidlington,Oxford,OX51GB,UK
225WymanStreet,Waltham,MA02451,USA
Firstpublished2014
Copyrightr2014ElsevierInc.Allrightsreserved
Nopartofthispublicationmaybereproduced,storedinaretrievalsystem,ortransmitted
inanyformorbyanymeanselectronic,mechanical,photocopying,recordingor
otherwisewithoutthepriorwrittenpermissionofthepublisher.
PermissionsmaybesoughtdirectlyfromElsevier’sScience&TechnologyRights,
DepartmentinOxford,UK:phone[phone];fax[phone];
email:[email protected],visittheScienceandTechnologyBooks
websiteatwww.elsevierdirect.com/rightsforfurtherinformation.
Notice
Noresponsibilityisassumedbythepublisherforanyinjuryand/ordamagetopersons,
orpropertyasamatterofproductsliability,negligenceorotherwise,orfromanyuseor
operationofanymethods,products,instructionsorideascontainedinthematerialherein.
Becauseofrapidadvancesinthemedicalsciences,inparticular,independent
verificationofdiagnosesanddrugdosagesshouldbemade.
BritishLibraryCataloguing-in-PublicationData
AcataloguerecordforthisbookisavailablefromtheBritishLibrary
LibraryofCongressCataloging-in-PublicationData
AcatalogrecordforthisbookisavailablefromtheLibraryofCongress
ISBN:978-0-12-401689-7
ForinformationonallAcademicPresspublications
visitourwebsiteatelsevierdirect.com
PrintedandboundinUnitedStatesofAmerica
14 15 16 17 10 9 8 7 6 5 4 3 2 1
Landon and Mason
A continuous source of joy and inspiration
And a reminder to keep asking why why why....
Preface
Ifweknewwhatitwasweweredoing,itwouldnotbecalledresearch,wouldit?
AlbertEinstein
The Science of Algorithmic Trading and Portfolio Management is a reference book intended
to provide traders, portfolio managers, analysts, students, practitioners, and financial execu-
tives withan overview of the electronic trading environment, and insight intohow algorithms
canbeutilizedtoimproveexecutionqualityandfundperformance.
We provide a discussion of the current state ofthe marketand advanced modeling techniques
fortradingalgorithms,stockselection,andportfolioconstruction.
Thisreferencebookwillprovidereaderswith:
(cid:1) Anunderstandingofthenewelectronictradingenvironment.
(cid:1) Anunderstandingoftransactioncostanalysis(TCA)andpropermetricsforcostmeasurement
andperformanceevaluation.
(cid:1) A thorough understanding of the different types of trading algorithms: liquidity seeking,
dark pools, arrival price, implementation shortfall (IS), volume weighted average price
(VWAP),arrivalprice,andportfolioimplementationshortfall.
(cid:1) Provenmarketimpactmodelingtechniques.
(cid:1) An understanding of algorithmic trading across various asset classes: equities, futures,
fixedincome,foreignexchange,andcommodities.
(cid:1) Advancedalgorithmicforecastingtechniquestoestimatedailyliquidityandmonthlyvolumes.
(cid:1) An algorithmic decision making framework to ensure consistency between investment
andtradingobjectives.
(cid:1) Abestexecutionprocess.
Readerswillsubsequentlybepreparedto:
(cid:1) Developreal-timetradingalgorithmscustomizedtospecificinstitutionalneeds.
(cid:1) Designsystemstomanagealgorithmicriskanddarkpooluncertainty.
(cid:1) Evaluate market impact models and assess performance across algorithms, traders, and
brokers.
(cid:1) Implementelectronictradingsystems.
For the first time, portfolio managers are not forgotten and will be provided with proven
techniquestobetterconstructportfoliosthrough:
(cid:1) StockSelection
(cid:1) PortfolioOptimization
(cid:1) AssetAllocation
(cid:1) MIFactorScores
xv
xvi Preface
(cid:1) Multi-AssetInvesting
(cid:1) FactorExposureInvesting
The book is categorized in three parts. Part I focuses on the current electronic market envi-
ronment where we discuss trading algorithms, market microstructure research, and transac-
tion cost analysis. Part II focuses on the necessary mathematical models that are used to
construct, calibrate, and test market impact models, as well as to develop single stock and
portfolio trading algorithms. The section further discusses volatility and factor models, as
well as advanced algorithmic forecasting techniques. Part III focuses on portfolio manage-
ment techniques and how TCA and market impact can be incorporated into the investment
decisions, stock selection, and portfolio construction to improve portfolio performance. We
introduce readers to an advanced portfolio optimization process that incorporates market
impact and transaction costs directly into the portfolio optimization. We provide insight into
howMIfactorscorescanbeusedtoimprovestockselection,aswellasatechniquethatcan
beusedbyportfoliomanagerstodecipherbroker-dealerblackboxmodels.Thissectioncon-
cludes with an overview of high frequency trading, and the necessary mathematical knowl-
edgerequiredtodevelopblackboxtradingmodels.
Acknowledgments
There are several people who made significant contributions to the concepts introduced
throughout the text. Without their insights, comments, suggestions, and criticism, the final
versionofthisbookandthesemodelswouldnothavebeenpossible.Theyare:
Roberto Malamut, Ph.D., was instrumental in the development of the methodologies and
framework introduced in this book. His keen mathematical insight and market knowledge
helpedadvancemanyofthetheoriespresentedthroughoutthetext.MortonGlantz,mycoau-
thor from Optimal Trading Strategies, provided invaluable guidance and direction, and
helped turn many of our original ideas into formulations that have since been put into prac-
ticebytradersandportfoliomanagers,andhavenowbecomemainstreamintheindustry.
The All-Universe Algorithmic Team: Roberto Malamut (again), Andrew Xia, Hernan Otero,
DeepakNautiyal,DonSun,KevinLi,PeterTannenbaum,ArunRajasekhar,andMustaqAli,
and Tom M. Kane and Dan Keegan too! And to complete the All-Universe team: Pierre
Miasnikof, Agustin Leon, and Alexis Kirke for all of their early contribution in developing
and testing many of the ideas and modelsthat havenow become ingrained into the algorith-
mictradinglandscape.Theircontributiontoalgorithmictradingissecondtonone.
Wayne Wagner provided valuable direction and support over the years. His early research
has since evolved into its own science and discipline known as transaction costs analysis
(TCA).Hisearlyvisionandresearchhashelpedpavethewayformakingourfinancialmar-
kets more efficient and investor portfolios more profitable. Robert Almgren and Neil Chriss
provided the ground breaking work on the efficient trading frontier, and introduced the
appropriate mathematical trading concepts to the trading side of the industry. Their seminal
paper on Optimal Liquidation Strategies is the reason that trading desks have embraced
mathematicalmodelsandalgorithmictrading.
Victoria Averbukh Kulikov, Director of Cornell Financial Engineering Manhattan (CFEM),
allowed me to lecture on Algorithmic Trading (Fall2009 & Fall 2010) and test many of my
theories andideas ina classsetting.I haveagreatdeal of gratitude toherand toallthe stu-
dents for correcting my many mistakes before they could become part of this book. They
providedmoreanswerstomethanIamsureIprovidedtothemduringthesemester.
Connie Li, Quantitative Analyst at Numeric Investments (and M.S. in Financial Engineering
from Cornell University), provided invaluable comments and suggestions throughout the
writing of the book. And most importantly, corrected the errors in my math, the grammar in
my writing, and helped simplify the many concepts discussed throughout the book. Scott
Wilson, Ph.D., Analyst at Cornerstone Research, provided invaluable insight and direction
for modeling trading costs across the various asset classes, and was influential in helping to
structuretheconceptsbehindthefactorexposureallocationscheme.
xvii
xviii Acknowledgments
Ayub Hanif, Ph.D. Researcher, Financial Computing and Applied Computational Science,
University College London, for his extraordinary contribution to the book as the author of
Chapter 13: High Frequency Trading and Black Box Models. This chapter has provided
more insight into the secretive word of black box modeling and high frequency trading than
has been disseminated in all the seminars and conferences I have attended put together. It is
amustreadforanyinvestorseekingtomanageaportfolioandearnaprofitintheultracom-
petitivehighfrequencyandhighvelocitytradingspace.
Additionally, Dan Dibartolomeo, Jon Anderson, John Carillo, Sebastian Ceria, Curt Engler,
Marc Gresack, Kingsley Jones, Scott Wilson, Eldar Nigmatullin, Bojan Petrovich, Mike
Rodgers, Deborah Berebichez, Jim Poserina, Mike Blake, and Diana Muzan for providing
valuable insight, suggestions, comments, during some of the early drafts of this manuscript.
This has ultimately lead to a better text. The team at Institutional Investor and Journal of
Trading,AllisonAdams,BrianBruce,andDebraTraskforongoingencouragementandsup-
portontheresearchsideofthebusiness.
A special thanks to Richard Rudden, Stephen Marron, John Little, Cheryl Beach, Russ
Feingold,KevinHarper,WilliamHederman,JohnWile,andKyleRudden,frommyfirstjob
out of college at R.J. Rudden Associates (now part of Black and Veatch) for teaching the
true benefits of thinking outside of the box, and showing that many times a non-traditional
approachcouldoftenprovetobethemostinsightful.
Finally, Hans Lie, Richard Duan, Trista Rose, Alisher Khussainov, Thomas Yang, Joesph
Gahtan,FabienneWilmes,ErikSulzbach,CharlieBehette,MinMoon,KapilDhingra,Harry
Rana, Michael Lee, John Mackie, Nigel Lucas, Steve Paridis, Thomas Reif, Steve Malin,
Marco Dion, Michael Coyle, Anna-Marie Monette, Mal Selver, Ryan Crane, Matt Laird,
Charlotte Reid, Ignor Kantor, Aleksandra Radakovic, Deng Zhang, Shu Lin, Ken Weston,
Andrew Freyre-Sanders, Mike Schultz, Lisa Sarris, Joe Gresia, Mike Keigher, Thomas
Rucinski, Alan Rubenfeld, John Palazzo, Jens Soerensen, Adam Denny, Diane Neligan,
Rahul Grover, Rana Chammaa, Stefan Balderach, Chris Sinclaire, James Rubinstein, Frank
Bigelow, Rob Chechilo, Carl DeFelice, Kurt Burger, Brian McGinn, Dan Wilson, Kieran
Kilkenny, Kendal Beer, Edna Addo, Israel Moljo, Peter Krase, Emil Terazi, Emerson Wu,
Trevor McDonough, Simon, Jim Heaney, Emilee Deutchman, Seth Weingram, and Jared
Anderson.
BestRegards,
RobertKissell,Ph.D.
1
Chapter
Algorithmic Trading
INTRODUCTION
Algorithmic trading represents the computerized executions of financial
instruments. Algorithms trade stocks, bonds, currencies, and a plethora
of financial derivatives. Algorithms are also fundamental to investment
strategies and trading goals. The new era of trading provides investors
with more efficient executions while lowering transaction costs—the
result, improved portfolio performance. Algorithmic trading has been
referredtoas“automated,”“blackbox”and“robo”trading.
Trading via algorithms requires investors to first specify their investing
and/or trading goals in terms of mathematical instructions. Dependent
uponinvestors’needs,customizedinstructionsrangefromsimpletohighly
sophisticated.Afterinstructionsarespecified,computersimplementthose
tradesfollowingtheprescribedinstructions.
Managersusealgorithmsinavarietyofways.Moneymanagementfunds—
mutual and index funds, pension plans, quantitative funds and even hedge
funds—use algorithms to implement investment decisions. In these cases,
money managers use different stock selection and portfolio construction
techniques to determine their preferred holdings, and then employ algo-
rithmstoimplementthosedecisions.Algorithmsdeterminethebestwayto
slice orders and trade over time. They determine appropriate price, time,
and quantity of shares (size) to enter the market. Often, these algorithms
makedecisionsindependentofanyhumaninteraction.
Similartoamoreantiquated,manualmarket-makingapproach,brokerdealers
andmarketmakersnowuseautomatedalgorithmstoprovideliquiditytothe
marketplace. As such, these parties are able to make markets in a broader
spectrum of securities electronically rather than manually, cutting costs of
hiringadditionaltraders.
Aside from improving liquidity to the marketplace, broker dealers are
usingalgorithmstotransactforinvestorclients.Onceinvestmentdecisions
are made,buy-sidetrading deskspassorderstotheir brokers for execution
TheScienceofAlgorithmicTradingandPortfolioManagement.DOI:http://dx.doi.org/10.1016/B978-0-12-401689-7.00001-5
©2014RobertKissell.PublishedbyElsevierInc.Allrightsreserved. 1
2 CHAPTER 1 Algorithmic Trading
using algorithms. The buy-side may specify which broker algorithms to
use to trade single or basket orders, or rely on the expertise of sell-side
brokers to select the proper algorithms and algorithmic parameters. It is
importantforthesell-sidetopreciselycommunicatetothebuy-sideexpec-
tationsregardingexpectedtransactioncosts(usuallyvia pre-trade analysis)
and potential issues that may arise during trading. The buy-side will need
to ensure these implementation goals are consistent with the fund’s
investment objectives. Furthermore, itiscrucial forthe buy-sidetodeter-
mine future implementation decisions (usually via post-trade analysis) to
continuously evaluate broker performance and algorithms under various
scenarios.
Quantitative, statistical arbitrage traders, sophisticated hedge funds, and
the newly emerged class of investors known as high frequency traders
will also program buying/selling rules directly into the trading algorithm.
The program rules allows algorithms to determine instruments and how
theyshould beboughtandsold.Thesetypes ofalgorithms arereferred to
as“blackbox”or“profitandloss”algorithms.
For years, financial research has focused on the investment side of a busi-
ness. Funds have invested copious dollars and research hours on the quest
for superior investment opportunities and risk management techniques,
with very little research on the implementation side. However, over the
last decade, much of this initiative has shifted towards capturing hidden
value during implementation. Treynor (1981), Perold (1988), Berkowitz,
Logue, and Noser (1988), Wagner (1990), and Edwards and Wagner
(1993) were among the first to report the quantity of alpha lost during
implementation of the investment idea due to transaction costs. More
recently, Bertsimas and Lo (1996), Almgren and Chriss (1999, 2000),
Kissell, Glantz, and Malamut (2004) introduced a framework to minimize
market impact and transaction costs, as well as a process to determine
appropriateoptimalexecutionstrategies.Theseeffortshavehelpedprovide
efficientimplementation—theprocessknownasalgorithmictrading1.
While empirical evidence has shown that when properly specified, algo-
rithms result in lower transaction costs, the process necessitates investors
be more proactive during implementation than they were previously uti-
lizing manual execution. Algorithms must be able to manage price, size,
and timing of the trades, while continuously reacting to market condition
changes.
1A review of market microstructure and transaction cost literature is provided in
Chapter2,MarketMicrostructure.