Seminar (San Francisco) - Steven Pav - Portfolio Inference and Portfolio Overfit

Itinerary:

6pm: Reception in Julia's Lounge

7pm: Talk in the Member's Lounge

8pm: Networking*

Full title: Portfolio Inference and Portfolio Overfit

Abstract: By using a little known matrix equation, we derive the asymptotic distribution of the Markowitz portfolio, taking into account common practical tweaks. This allows inference to be performed in a wide variety of asset allocation problems. We also discuss a fundamental bound on portfolio quality when the goal is to maximize the Sharpe ratio. This bound, which can be seen as a quantification of 'overfit', helps solve some puzzles in asset allocation: why diversification can hurt, why portfolio managers do not (or should not) make more than 5 decisions at a time, etc.

About the Speaker: Steven Pav served as a quantitative strategist at Cerebellum Capital for 7 years where he designed and implemented backtest, execution, and research infrastructure in Matlab and C for a daily trading system on equities and volatility futures. His contributions also include designing machine learning quantitative strategies and devising methods to correct for overfit bias in the backtesting and strategy development process. Steve holds a Ph.D. in Math from Carnegie Mellon.

About the Thalesians: The Thalesians are a think tank of dedicated professionals with an interest in quantitative finance, economics, mathematics, physics and computer science, not necessarily in that order. Learn more about us at http://www.thalesians.com .

Acknowledgements: The Thalesians are delighted to acknowledge the sponsorship of this event by Voleon Capital Management.

• Please note the bar in Julia's Lounge will remain open to the public until 8:30pm, but the Member's Lounge will remain open to the Thalesians until 9:30pm.