Thalesians/QFGG (Frankfurt) - Thomas Wiecki - Predict out-of-sample performance

NOTE: the event is at 6.30pm Frankfurt time (please ignore the Outlook invite time)

Full title: Thomas Wiecki - Predicting out-of-sample performance and building multi-strategy portfolios using Random Forests

Thanks for Jochen Papenbrock and Adrian Zymolka for organising and for PPI AG for hosting. Tickets will be FREE for this event!

You can access the Thalesians/Quant Finance Germany (Frankfurt) LinkedIn Group page here (https://www.linkedin.com/grp/home?gid=8321682).

Abstract

The question of how predictive a backtest is of out-of-sample performance is at the heart of algorithmic trading. Using a unique dataset of 888 algorithmic trading strategies developed and backtested on the Quantopian platform with at least 6 months of out-of-sample performance, we study the prevalence and impact of backtest overfitting. Specifically, we find that commonly reported backtest evaluation metrics like the Sharpe ratio offer little value in predicting out of sample performance (R² < 0.025). However, we show that by training a Random Forest regressor on a variety of features that describe backtest behavior, out-of-sample performance can be predicted at a much higher accuracy (R² = 0.17) on hold-out data compared to using linear, univariate features. We then show that we can construct a multi-strategy portfolio based on predictions by the Random Forest which performed significantly better out-of-sample than other alternatives.

Speaker: Thomas Wiecki is the Data Science Lead at Quantopian focusing Bayesian models to evaluate trading algorithms. Previously, he was a Quantitative Researcher at Quantopian developing an open-source trading simulator as well as optimization methods for trading algorithms.

Thomas holds a PhD from Brown University.