Noon Barbari
Registrati
Backtesting

Walk-forward optimization

Optimize on a rolling in-sample window, validate on the next out-of-sample slice.

Walk-forward optimization (WFO) is the standard cure for overfitting in parameter tuning. The historical period is split into a sequence of windows; for each window, the parameters are optimized on the in-sample portion and then evaluated on the immediately following out-of-sample portion. The window then rolls forward and the process repeats.

The result is a stitched-together out-of-sample equity curve: every trade in the OOS curve was generated by parameters that had not seen its bar. This is much closer to what live trading would have produced than a single optimized backtest over the whole history.

WFO is parameter-set sensitive: optimizing 12 hyperparameters on 6 months of data still overfits, walk-forward or not. The discipline is about how you split data, not a free pass to search a giant parameter space.

Come Noon Barbari usa Walk-forward optimization

Ogni concetto qui è implementato nella piattaforma. Apri la documentazione o lo strumento corrispondente per vederlo all'opera.

Walk-forward in the platform

Termini correlati

Torna al glossario