Skip to content

Cross-validation

When trading strategies that have shown strong performance in backtesting are applied to new data, they often prove to be fragile or may even fail catastrophically. This usually happens due to backtest overfitting, which can be detected and avoided with proper cross-validation.

  • We will start with manual cross-validation, then cover different splitting schemes and techniques for analyzing split distributions. Finally, we will go over how to automate various cross-validation steps, from splitting regular arrays and complex VBT objects to testing entire parameter-based and ML-based pipelines 🔥
  • Example: Median Sharpe by market regime


View