A bayesian approach to measurement of backtest overfitting

Quantitative investment strategies are often selected from a broad class of candidate models estimated and tested on historical data. Standard statistical techniques to prevent model overfitting such as out-sample backtesting turn out to be unreliable in situations when the selection is based on results of too many models tested on the holdout sample. There is an ongoing discussion of how to estimate the probability of backtest overfitting and adjust the expected performance indicators such as the Sharpe ratio in order to reflect properly the effect of multiple testing. We propose a consistent Bayesian approach that yields the desired robust estimates on the basis of a Markov chain Monte Carlo (MCMC) simulation. The approach is tested on a class of technical trading strategies where a seemingly profitable strategy can be selected in the näive approach.