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.
Comment Noon Barbari utilise Walk-forward optimization
Chaque concept ici est implémenté dans la plateforme. Ouvre la documentation ou l'outil concerné pour le voir en action.
Walk-forward in the platform →Termes liés
- Backtesting
Backtest
Simulating a trading rule on historical data to estimate how it would have performed.
- Backtesting
Out-of-sample
Data the strategy was not allowed to see during parameter selection.
- Backtesting
In-sample
The portion of history used to fit the strategy's parameters.
- Backtesting
Overfitting
Fitting a strategy so closely to past data that it captures noise, not signal.