Docstoc

Download - OptiRisk Systems

Document Sample
Download - OptiRisk Systems Powered By Docstoc
					               Challenges for algorithmic
                       execution
                                Gordon Baker




2007-06-26 CARISMA conference                  1
Agenda

• Context
• Research
• To do




2007-06-26 CARISMA conference   2
Some definitions
Execution vs. trading

• Execution
      – No choice of asset, direction or quantity
      – Usually intra-day or good for day orders
• Trading … everything else
      – Market making
      – Arbitrage
      – Model based trading


2007-06-26 CARISMA conference                       3
A brief history
1980s

Technique                       Significance
• VWAP, TWAP and                • Market structure (electronic
   participation strategies        order books) determines
• Empirical approaches to          spread of automated
   market impact                   strategies




2007-06-26 CARISMA conference                                4
A brief history
1990s

Technique                       Significance
• Bertsimas & Lo 1998           • ITG is first algorithmic
• Further gaming                   brokerage
   countermeasures              • 1997-2000 day trading:
                                   everyone became a model
                                   driven trader




2007-06-26 CARISMA conference                                5
A brief history
2000s

Techniques                          Significance
• Landmark research                 • Credit Suisse launches AES
      – Almgren & Chriss 2000          in 2000
      – Almgren 2001                • DMA and algorithmic
      – Kissell, Glantz & Malamut      execution account for 50%
        2003                           of NYSE turnover
      – Kissell & Malamut 2005
                                    • Some broker-dealers have
• Arrival price strategies             80% of order flow executed
• Portfolio strategies                 by algorithmic strategies



2007-06-26 CARISMA conference                                       6
Algorithmic execution strategies
A broad classification

Completing                           Non-completing
• Quantities                         • Participation
      – VWAP                         • Relative value
      – TWAP
• Participation rates
      – Arrival price or shortfall




2007-06-26 CARISMA conference                           7
Where we are now
“Algorithms are only good for liquid stocks”

Broker-dealers                  Investors
• Widespread use of             • Broker-dealers algorithms
  algorithmic execution,           used for 15-30% of orders
  either directly or when       • Increasing addition of
  unwinding                        constraints to orders
• Brand explosion of            • Consequence: completing
  strategies                       strategies become non-
                                   completing




2007-06-26 CARISMA conference                                  8
Execution schedules
Research has provided the benchmarks
                         8%
                                                                                              TWAP
                                                                                              VWAP
                                                                                              VWAP exp -3t
                                                                                              VWAP exp -6t
                         7%




                                                                                              cosh T-t
                                                                                              cosh 3(T-t)
                                                                                              cosh 6(T-t)
                         6%
                         5%
       relative volume

                         4%
                         3%
                         2%
                         1%
                         0%




                              08:00   09:00   10:00   11:00   12:00   13:00   14:00   15:00        16:00



2007-06-26 CARISMA conference                                                                                9
Execution models

Strengths                       Weaknesses
• Mathematically sound          • Assume perfect markets
   approaches to optimising     • Do not address tactics or
   variance                       order placement decisions
• Common ground for
   investors and broker-
   dealers




2007-06-26 CARISMA conference                                 10
Other academic contributions

• Econophysics
      – Models for non-gaussian returns
      – Minority games
• Behavioural finance




2007-06-26 CARISMA conference             11
   Challenge #1

   VOLUME


2007-06-26 CARISMA conference   12
The extremes for algorithmic execution
Volume and trade frequency within the FT-SE 100
                                                       PRTY .L



                                                                                            SHP.L
                                                1.0
       std dev relative to mean, daily volume




                                                                       BI.L
                                                                       Y ELL.L


                                                                                           BGY .L
                                                0.8




                                                                    FP.L          SN.L
                                                                                          IMT.L
                                                                    EXPN.L CPG.L
                                                                        DSGI.L
                                                                    IAP.L
                                                                     LII.L       CNE.L
                                                                                  SPW.L
                                                                                 MKS.L                       GUS.L
                                                                    ITV.L    HNS.L
                                                                    DRX.L           ULVR.L VED.L                                                                 AZN.L
                                                0.6




                                                                           CW.L
                                                                       WTB.L
                                                      SL.L                 SGE.L                                                             HSBA.L
                                                                                    WPP.L
                                                                  REX.L           SBRY .L
                                                                                     ICI.L
                                                                               BSY
                                                                            BB.L .L                       PRU.L                           BARC.L
                                                                                                   RB.L
                                                                     CPI.L    REL.L           NG.L
                                                                                 AVZ.L
                                                                              GLH.L            BAY .L              LLOY .L                               GSK.L
                                                                                    OML.L
                                                                                    LGEN.L
                                                                     III.L NRK.LTATE.LLMI.LRR.L
                                                                     ETI.L
                                                                    PSN.L                                                        CS.L
                                                                  AL.L SCTN.L  SVT.L ANTO.L EMG.L
                                                                           MRW.L     RTR.L CBRY .L BA.L
                                                                                       CCL.L
                                                                                      KGF.L                                                 RBS.L
                                                                                          NXT.L
                                                                                             IPR.L
                                                                                              WOS.L             BG.L
                                                                                 SMIN.L SSE.L
                                                                                         SAB.L
                                                0.4




                                                                  HOME.L
                                                             RSL.L ABF.L KEL.L                           HBOS.L
                                                                           PSON.L                          BT.L                                         XTA.L
                                                                  KAZ.L         IHG.L                             RDSb.L
                                                                                    RDSa.L                       TSCO.L
                                                                        RSA.L
                                                                                      BLND.LCNA.L STAN.L                          VOD.L
                                                                                       UU.L           BATS.L
                                                                    HMSO.L
                                                             SLOU.L JMAT.L                              DGE.L                                          AAL.L
                                                                                                       AV.L                                                              BLT.L   RIO.L
                                                                                  LAND.L                                                                                  BP.L




                                                                   2                              4                          6                     8                        10

                                                                                                         mean trades per minute
2007-06-26 CARISMA conference                                                                                                                                                            13
The well behaved …
High trading frequency – low volume variance
                                  2006-10                2006-11   2006-12   4%


                                                                             3%


                         AZN.L                                               2%


                                                                             1%


                                                                             0%
                         BLT.L
                         BP.L




                         08:00   11:00   14:00   16:30


2007-06-26 CARISMA conference                                                     14
… and the not so well behaved
Low trading frequency – high volume variance
                                   2006-10                2006-11   2006-12   4%


                                                                              3%


                         BGY.L                                                2%


                                                                              1%


                                                                              0%
                         SHP.L
                         YELL.L




                         08:00    11:00   14:00   16:30


2007-06-26 CARISMA conference                                                      15
Even the more liquid …
BP.L until mid November 2006
                         6%
                                                                                          mean
                                                                                          standard deviation
                         5%
                         4%
       relative volume

                         3%
                         2%
                         1%
                         0%




                              08:00   09:00   10:00   11:00   12:00   13:00   14:00   15:00    16:00



2007-06-26 CARISMA conference                                                                                  16
… can surprise
BP.L traded twice its mean daily volume on November 17
                         6%
                                                                                          mean
                                                                                          standard deviation
                                                                                          2006-11-17
                         5%
                         4%
       relative volume

                         3%
                         2%
                         1%
                         0%




                              08:00   09:00   10:00   11:00   12:00   13:00   14:00   15:00    16:00



2007-06-26 CARISMA conference                                                                                  17
Coping
Towards a common understanding of behaviour

• Median curves?
• Volatile daily volume has no effect for
  completing strategies constrained by time
• Prediction of relative weighting?




2007-06-26 CARISMA conference                 18
   Challenge #2

   EXECUTION MICRO-ECONOMICS


2007-06-26 CARISMA conference   19
Order submission
The fundamental state model is simple




                          LIM   WAIT    FAK




2007-06-26 CARISMA conference                 20
Order submission
Some of the microeconomics

                    Price making       Price taking
             10
             8
       bps

             6
             4




                                                                   volatility
                                                                   spread
                                                                   best size cleared
                                                                   volatility exceeds spread


                        2          4   6          8      10   12              14

                                               minutes
2007-06-26 CARISMA conference                                                                  21
   SUMMARY


2007-06-26 CARISMA conference   22
Algorithmic execution

• One of the major execution methods
• Contribution from research is clear
• More is needed to keep the mutual
  expectations of buy- and sell-side in line
• Two challenges today
      – Turning attention on the volume dimension
      – Formalising the different levels of execution
        model

2007-06-26 CARISMA conference                           23
                     Algorithmic execution

                         Data provided by
                            Analysis tools



                    Further information      http://www.gbkr.com

2007-06-26 CARISMA conference                                      24

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:2
posted:8/24/2012
language:English
pages:24