Historical Backtesting vs. Real-World Positions

1
Historical Backtesting vs. Real-World Positions
George R. Brown School of Engineering
STATISTICS

• SECOND EUBANK CONFERENCE MODELING FINANCIAL
  MARKETS IN A WORLD OF FIAT MONEY
• John A. Dobelman
• Rice University
• October 18-19, 2010

2
(No Transcript)
3
Outline

• The Asset Allocation Problem
• The Equity Portfolio
• Managing the Portfolio
• Conclusion and Future Work

3
4
Asset Allocation
WSJ, October 19, 2030
5
Simplified Allocation

• October 14, 2009 thru October 15, 2010
• WTC Insurance
• 32,536 6-Yr Return
• Gold (Comex)
• 28.2 1-Year Return
• Cotton (ICE)
• 63.5 1-Year Return
• TGMP (American Power Group)
• 850 1-Yr Return
• LBSV (Liberty Silver)
• 9,499,900 1-Year Return

6
Best Allocation
LBSV Liberty Silver, OTC-BB 10/14/09 10/15/10

• Powerball (50M Jackpot)
• 4,999,999,900 1-Day return

7
Market Modeling

• Primarily for Trading
• Determine how much to produce/buy
• Capacity allocation
• Hedging application
• Speculation!
• Not Necessary Necessary for LT Holding
• LT holding returns ? GBM model
• Given that, looking for w(w1,,wk)

8
BAH ? BAH

• Active portfolio management required
• "Indexes are not available for direct investment
  therefore, their performance does not reflect the
  expenses associated with the management of an
  actual portfolio. -DFA
• Even the E-V portfolios require rebalancing to
  maintain MV construction
• For
• SBAH BAH DivMgt
• SBAH BAH DivMgt Tender
• SBAH BAH DivMgt Tender Tax
• SBAH BAH DivMgt Tender Tax Realloc

9
Come a Long Way Since 1973
10
ACC2010

• 2010 American Control Conference
• Operations to Finance Opportunities for Control
  Theory and Application
• Control Systems Methods in Finance Modeling and
  Optimal Trading, Primbs, Stanford University, and
  Barmish, Univ. Wisconsin
• Interfaces between control theory and finance
• Dynamic hedging as a stochastic control problem
• LQ and receding horizon control methodolgies

11
Thats not All!

• A model of the human as a suboptimal smoother
• WB Rouse - 1974 IEEE Conference on Decision and
  Control, 1974
• Trading Costs Around MA Announcements
• L Mai, BF Van Ness, RA Van Ness, 1983
• Economic prediction using neural networks The
  case of IBM daily stock returns
• H White - Proceedings of the IEEE International
  Conference on , 1988
• Applications of statistical physics to economic
  and financial topics
• M Ausloos, N Vandewalle, P Boveroux, A - Physica
  A Statistical , 1999

12
1990s - 2010

• Chaos in futures markets? A nonlinear dynamical
  analysis (1991)
• Steven C. Blank, Journal of Futures Markets
• Components of multifractality in high-frequency
  stock returns (2005)
• J Kwapien Physica A Stat Mech Apps
• A fuzzy control model (FCM) for dynamic portfolio
  management
• R Östermark Fuzzy sets and Systems 1996
• Fluctuations and Market Friction in Financial
  Trading
• Bernd Rosenow, 2001, Condensed Matter

13
1990s - 2010

• Stochastic Lotka-Volterra Systems of Competing
  Auto-Catalytic Agents Lead Generically to
  Truncated Pareto Power Wealth Distribution,
  Truncated Levy Distribution of Market Returns,
  Clustered Volatility, Booms and Crashes
• Sorin Solomon (Hebrew University) Submitted on 30
  Mar 1998) Computational Finance 97
• THE JOINT PRICING OF VOLATILITY AND LIQUIDITY!
• F. Bandi, C.E. Moise, and J. Russell,2008
• Liquidity skewness
• R Roll, A Subrahmanyam - Journal of Banking
  Finance, 2010

14
1990s - 2010

• Idiosyncratic Volatility, Stock Market
  Volatility, and Expected Stock Returns
• Hui Guo, Robert Savickas. Journal of Business and
  Economic Statistics. January 1, 2006
• A theory of power-law distributions in financial
  market fluctuations
• X Gabaix, P Gopikrishnan, et.al. Nature 423
  (2003)
• On fitting the ParetoLevy distribution to stock
  market index data Selecting a suitable cutoff
  value
• H.F. Coronel-Brizioa, and A.R. Hernández-Montoya,
  Physica A Statistical Mechanics and its
  Applications Volume 354, 15 August 2005

15
1990s - 2010

• Predicting Stock Prices Using a Hybrid Kohonen
  Self Organizing Map (SOM)
• Afolab Olude System Sciences, 2007. HICSS
  2007
• Examples of these methods are fuzzy logic, neural
  network and hybridized methods such as hybrid
  Kohonen self organizing map (SOM), adaptive
  neuro-fuzzy inference system (ANFIS) etc.
• This paper presents a number of methods used to
  predict the stock price of the day. These methods
  are backpropagation, Kohonen SOM, and a hybrid
  Kohonen SOM...the hybrid Kohonen SOM is a better
  predictor compared to Kohonen SOM and
  backpropagation

16
Orthodoxy

• Departures from the EMH Market Portfolio
• Market OO
• Departure 1 OOE
• Departure 2 OOE\Priv
• Departure 3 OOS
• Departure 4 OOIndex
• Departure 5 O?Your E-V portfolio, m and s
• Departure 6 O?Your E-V portfolio,
• Departure 7 O? Some other portfolio P

17
Portfolio Construction

• Remark OOindex
• Wilshire 5000, SP500, RUT3000, Value-Line, DOW30,
  etc., are ALL actively determined portfolios.
• Only recently could you buy into a mutual
  fund/ETF which attempts to replicate these
  indexes
• Unless you inherit a portfolio, you must create
  one, or build one over time.

18
Portfolio Construction

• Fundamental analysis
• Slow and time-consuming
• Technical Analysis
• Value Line
• ONeil /Investors Business Daily
• Efficacy in question
• Quantitative Portfolio Management
• Formulation
• Management
• Allows statistics-based portfolio strategies

19
Portfolio Formulation

• You Must Pick 10 Stocks

20
How Much Would You Pay?
21
For This?
22
Fundamental Analysis
23
Fundamental Analysis
24
Fundamental Approach
25
Outliers/Outliars
26
BAH with the Greats

• Benjamin Graham
• Criteria for Defensive Investor 12/31/70
• Size 100M sales (326M today)
• Financial Strength CR 21, LTDltWC
• Positive EPS in last 10 years
• 20 years of uninterrupted dividends
• Min 33 EPS growth in 10 years
• PE lt 15 for last 3 years average EPS
• P/BV lt 15-22

27
Graham Portfolio

• As of 12/31/1970, this was the portoflio
• AC, American Can
• T, ATT
• A, Anaconda
• SWX, Swift
• Z, F.W. Woolworth
• Bring up to the present
• 1/4/1971 - End

28
Graham Portfolio
29
Graham BAH Results
CAGR from 1971 Thru 12/31/2009 4.13 Thru
10/10/2010 4.09 Original DOW 30 from
1971 Components unchanged 1956-1976 5 gone, only
14 continuously traded Original DOW 30 from
1/3/2000 Thru 10/10/2010 4.59 Indexes from
1/3/2000 thru 10/10/10 DJIAK -0.29 DOXIK
2.04 SP50 -2.03 SP50.R -0.24
30
10-Yr SP 500 Returns
31
30-Yr SP 500 Returns
32
Horizon Dependence
33
Benchmark Summaries
34
Benchmarks

• 50-Year Real Returns of 7 (Siegel, 2002)
• 1802 1870 (Schwert)
• 1871 1925 (Cowles)
• 1926 2001 (CRSP, all NYSE/AMEX/NASD)
• Post WWII 1946 2001
• Most inflation has been during this period

35
Benchmarks
36
Benchmarks
37
Return by Period
38
QPM

• Anomalies Research
• Ripe with Outperformance Goal
• Market Outperformance 433,000
• "Seeking Alpha 922,000
• Poor performance of mutual funds
• Quantitative Portfolio Management
• Matching market index
• Outperforming market index
• Beat the Index

39
QPM

• Characterized by lots of data
• Long look-back periods
• Backtesting
• Pitfalls
• Bad data
• Biases
• Datamining
• Transaction costs

40
Statistical QPM

• Lots and lots of quantitative funds
• Good job prospects, BTW E.g.,
• Quantitative Portfolio Analyst - Asset Manager
  for a Leading Hedge Fund
• Diversification and expansion has seen them
  create a traditional asset management fund.
• New York Up to 200k standard benefits and
  excellent bonus potential
• Options Strategy
• Public Domain
• Simugram

41
Time Value Option Sales
42
MaxMedian Rule
43
Simugram
44
Simugram
45
Fundamentals QPM

• Graham-Dodd on Steroids
• Exploit available data
• Try and sell for OPM
• Examples
• O'Shaughnessy
• Greenblatt
• Homegrown
• What happens in real life

46
James P. O'Shaughnessy

• c1920 Ignatius Aloysius O'Shaughnessy
• 110 million, I A O'Shaughnessy Foundation
• Avoided 20s stocks, fed his own companies
• 66 years gt 10M gt 5.4B (at 10)
• 1960 Jim O'Shaughnessy's Investment Horizon
  began
• 1986 BA Econ, University of Minnesota
• Began work at the family's VC firm
• 1988 O'Shaughnessy Capital Mgmt, Inc.
• Consulting to Institutional Investors

47
O'Shaughnessy (CONTD)

• 1995 Compustat (Standard Poor's)
• 1996 Cornerstone Growth and Value Funds
• 1997 "What Works on Wall Street, RBC
• 2000 Sold Cornerstone to Hennessey
• 200M Assets as of 6/30/00
• 2001 Sold O'Shaughnessy Capital to BSC
• About 500M
• 2005 Updated WWOWS

48
O'Shaughnessy (CONTD)

• 3Q2007 O'Shaughnessy Asset Mgmt, LLC
• Unwound in BSCM sale to JPM
• Taking 8B out BSAM's 44B
• Strategy
• Benchmark RUT2000
• No regard for sector
• Growth EPS Growth, 52W Price Increase, P/S
• Value Div Yield, LTM P/S , LTM P/CF

49
O'Shaughnessy (CONTD)

• Dreyfus Premier Alpha Growth Fund
• 1,600 companies
• 300 largest-market-cap
• 130 after P/E, 52W Price Incr, then by P/S
• Quarterly validation
• Dumping rules
• loss of 50 of market value
• takeover that doesn't meet the screens' criteria
• allegations of fraud
• bankruptcy

50
O'Shaughnessy-esgue

• Recall 11-year benchmarks

51
P/S Backtest
52
Starting and Stopping Times

• Backtested 1950 thru 1996 (47 years)
• Selected best factors (e.g., P/S)
• Started his funds
• Got lucky? In 4 years up about 166 vs. 75 for
  the market
• Got lucky? Sold to Hennessey at peak
• Got bad rap, 2000-2002 worse than market

53
HFCGX

• Hennessy Cornerstone Growth HFCGX
• 10 Year return 3.29
• Expenses 1.36
• Net 1.93
• DOW 1.7
• SP500 0.74
• Dow Div 4
• SP500 Div 2.5

54
HFCGX
55
Rules of the Game

• How much can your investors stomach
• Restrictions on redemptions
• Success begets loyalty
• Success depends on starting time

56
Another Example

• 10 year horizon
• Backtested 99-06, 12 fiscal periods (without
  drawdown constraint)
• Selected best factors (Screen1)
• Out-of-sample returns were still good
  (encouraging)
• Drawdowns horrible
• Developed Screen2 with drawdown constraint

57
Real Example, CONTD

• In 99-06 backtest, best return factor had
  drawdowns of
• Screen1 31
• Screen2 20
• In 99-09 OOS sample testing, best return factors
  remained the same (and equal IRRs of 33-34),
  but with drawdowns of
• Screen1 79.9
• Screen2 39

58
Out of Sample
59
Keeping in the Game
60
SYWTBAMM?

• Survival
• This is not an easy undertaking
• Survival depends on your starting time and
  redemption restrictions
• Young funds can have the least restrictive
  redemption requirements

61
A Special Thanks To

• Eubank Benefactors
• Profs. J.R. Thompson and E.E. Williams
• K.B. Ensor, Chair of CoFES
• TRU Dept. of Statistics
• Collaborators