Simulation as a Stock Market Backtesting Tool

Abstract

Backtesting stock market investment strategies is fraught with danger – for example, overfitting. The signal to noise ratio in stock markets is so low that overfitting is inevitable. Simulation offers a means of assessing and compensating for the dangers. It is not obvious at first how simulation can be helpful for backtesting and predicting markets but we show in five examples how this can be done. Techniques discussed are smoothing and regularization and are applied to investment strategies such as moving average crossover rules, momentum rules, and dynamic regression rules.

Keywords: Backtesting, Simulation, Smoothing, Data Mining Bias, Overfitting, Regularization, Moving Averages, Momentum, Alpha-Momentum

JEL Classification: C02, C15, C18, C52, C53, C63, G17