Future work on statistical significance
Make rigorous the data-snooping correction for evaluating rule accuracy.
Evaluate the joint significance of all rules in a strategy, accounting for the size of the search space.
Separate performance deterioration on a test set into components due to changes in market patterns, overfitting, and randomness.
Devise a sequential hypothesis-testing procedure (similar to a Shewhart chart for quality control) for recognizing when market conditions change, and when a strategy ceases to be useful.
Compare with the bootstrap method of Hal White for computing the p-value of a strategy given the fact many others have been evaluated.
- Data-Snooping, Technical Trading, Rule Performance and the Bootstrap by Ryan Sullivan, Allan Timmermann, and Halbert White, Journal of Finance, October 1999.