Performance also: WFA, walk-forward test, out-of-sample test

Walk-forward analysis

A validation method that repeatedly optimises EA parameters on a training window and tests on the immediately following unseen data window, mimicking real-world deployment conditions.

Why it matters

A standard backtest optimises parameters across the entire historical period, then measures performance on the same data. Walk-forward analysis separates optimisation and testing into separate windows, providing a genuine out-of-sample result.

Typical process

Optimise parameters on months 1-12.
Test on months 13-14 (out-of-sample).
Shift the window: optimise on months 3-14, test on 15-16.
Repeat across the full dataset.
Report the combined out-of-sample performance.

If out-of-sample performance is close to in-sample, the strategy is robust. Large gaps indicate overfitting.