EFTA00611842.pdf

Alex Gerasev Bounded Tradable Processes EFTA00611842 Confidentiality Information in this presentation is strictly private & confidential EFTA00611843 Background Strategy live at Buttonwood Group Trading (BGT) 07/2011-10/2012 (16 months); back-tested from 01/2006 with similar performance results Founder of Quanterra Advisors LLC and developer of quantitative trading strategies since 06/2010 Project Manager / Subject Matter Expert, Global Banking & Markets at RBS and ABN AMRO 2001-2010 M.Sc. in Financial Mathematics, University of Chicago B.Sc. in Physics & IT, Novosibirsk State University Research at VEPP-4 electron-positron collider of Novosibirsk Institute of Nuclear Physics (11 GeV energy range) EFTA00611844 Live Track Record Buttonwood Group Trading, 07/2011 — 10/2012 EFTA00611845 Live Track Record BG Date range 07/2011 —10/2012 (16 months) Reserved capital Average utilized capital (AUC) Available leverage Reinvestment of gains into the strategy No Asset classes 3.5m 1.4m 15x Equities (85%); futures (15%) o.97m (67.5% return on AUC) Positive months Negative months Best month Worst month Sharpe ratio 13 (81%) 3 (19%) o.27m -0.07M 2.67 EFTA00611846 Live Track Record - BGT Monthly Net PnL 300,000.00 250,000.00 200,000.00 150,000.00 100,000.00 50,000.00 0.00 -50,000.00 -100,000.00 list 2011-07 2011-08 2011-09 2011-10 2011-11 203.1-12 2012-01 2012-02 2012-04 is 2012-06 2012-07 2012-08 2012-09 2012-10 EFTA00611847 Live Track Record BGT Cumulative Net Return on AUC 80.00% 70.00% 6o.00% 5o.00% 40.00% 30.00% 20.00% 10.00% o.00% 2011-07 2011-08 2011-09 2011-10 2011-11 2011-12 2012-01 2012.02 2012.03 2012-04 2012-05 2012.06 2012.07 2012-08 2012-09 2012-10 EFTA00611848 Live Track Record — BGT Comments: Reinvestment of strategy gains back into the strategy would have resulted in significant performance improvement due to compounding (total return on AUC 89.9% vs. 67.5%) Strategy performance was negatively impacted on at least 3 occasions by emergency re-allocation of capital to other traders EFTA00611849 Back-Testing Track Record 01/2006 —12/2012 (84 months) EFTA00611850 Back-Testing Track Record Date range 01/2006 - 12/2012 (84 months) Reserved capital Average utilized capital (AUC) Available leverage Reinvestment of gains into the strategy No Asset classes 1.0M 0.62M 15X Same as live trading Positive months Negative months Best month Worst month Sharpe ratio 70 (83%) 14 (17%) 0.42M -0.17M 3.12 EFTA00611851 Back-Testing Track Record Observations: Slightly better performance vs. live performance during 07/2011 —10/2012 period (no capital re- allocation issues as mentioned above) Slightly better overall performance (higher average volatility) Market volatility results in better performance; best years for the strategy were 2008 and 2009 EFTA00611852 Strategy Description Bounded Tradable Processes EFTA00611853 Strategy Description: Bounded Tradable Processes "Process" can be single financial instrument (stock, future, option, FX, etc.) two-legged spread S = X— aY, where X and Y are financial instruments (stocks, futures, options, FX, etc.) three-legged spread S = X — aY— bZ, where X, Y, Z are financial instruments complex combination spreads e.g. S = aX/Y, S=aX —6Y/gZ, etc. "Tradable" means that all components of S can be traded at any time in required quantities (given liquidity constraints and other technicalities) EFTA00611854 Arbitrage Bounds Tradable Process that has lower, upper, or both lower and upper arbitrage bounds can be unwound at a pre- determined price Various "mechanics" enforcing bounds Asset swap ("fungibility"): X may be swapped for Y at holder's discretion — ensures that X > Y; if X < Y, buy bunch of X, swap for Y and sell Yfor higher price and riskless profit ("pure" arbitrage) Multi-step asset swap: similar to above, but involving multiple steps to unwind complex spread at pre- determined price "Hard" bounds — guaranteed to be able to unwind the process at a known price EFTA00611855 Non-Arbitrage Bounds Process should not go beyond certain values due to fundamentals; however there is no way to unwind if it does Example: "asset quality" Equity with voting rights vs. equity without voting rights Higher-grade commodity vs. lower-grade commodity Senior vs. subordinated debt Once the process moves beyond the bound and information spreads through the market, the relationship would typically get restored as participants buy superior asset for less, or holders of inferior asset replace it with superior asset while booking a profit More dangerous than arbitrage bounds — reflect in position sizing (e.g. io% of arbitrage-bounded trade) EFTA00611856 Example: Arbitrage Bounded Process ("hard" bound at zero) EFTA00611857 Example: Non-Arbitrage Bounded Process ("soft" bound at zero) EFTA00611858 Approach to Trading The closer the process gets to its bound, the bigger the position with l00% of target position size right at the bound As the process moves away from the bound, reduce position size As the process oscillates we generate PnL buy buying lower and selling higher The schedule of process values and corresponding position sizes is known as loading curve EFTA00611859 Loading Curve Example 00 1 10 40 70 30 7 10 7 20 • • I0 12 le 10 EFTA00611860 Profitability Analysis Volatility is good! Great opportunities when position can be entered or increased close to the bound and unwound as the process moves away Worst-case scenario: process approaching its bound with very low volatility and staying at the bound for prolonged period of time EFTA00611861 Other PnL Contributors ■ Market making • Intraday fluctuations ▪ Capturing market impact Mean-reversion EFTA00611862 Loading Curve Optimization Goal: create a loading curve shape that would maximize PnL minimize drawdowns minimize PnL volatility maximize Sharpe ratio other considerations Highly computationally-intensive optimization techniques based on historical data EFTA00611863 Competitive Advantage Space not nearly as crowded as pure arbitrage trades Requires significant R&D efforts that further reduce competition Unique and highly computationally intensive process of model calibration and optimization EFTA00611864 Capacity and Scalability EFTA00611865 Existing Product Groups 14 product groups (tradable processes) are fully set up and are available for trading Estimated capacity without adverse impact on performance: up to io —15m Further scalability with some performance impact: up to 20-25m EFTA00611866 Adding New Product Groups New product groups (tradable processes) may be added Approximately 2 months of R&D per product group Many interesting opportunities in international markets EFTA00611867 Expanding Across Asset Classes Same quantitative research methods may be applied to other asset classes and across asset classes Virtually unlimited scalability Some ideas: Commodity calendar spreads or spread options + physical storage Inter-commodity spreads + physical asset (refinery, power plant) Convertible debt Debt + CDS EFTA00611868 Summary EFTA00611869 Summary Time-tested strategy with excellent performance characteristics Market-neutral with no correlation with the S&P 500 Highly scalable Diversified portfolio of trades already live Robust quant research framework Large search universe for adding new product groups Further scalability by expanding across asset classes EFTA00611870