Walk-Forward Validation
Definition
A testing methodology where a strategy is optimized on historical data (in-sample) and then tested on unseen future data (out-of-sample) to verify it isn't overfit.
Example
A strategy is trained on 2022-2023 data, then tested on 2024 data. If it performs well on both, it passes walk-forward validation.
How Gilito AI Uses It
Every strategy undergoes walk-forward validation. Strategies that pass receive a 'Walk-Forward Validated' badge, indicating they're less likely to be overfit to historical data.