Walk-forward optimization

Walk-forward optimization is a way to test whether tuned parameters actually hold up, rather than just fitting the past. You optimize parameters on an in-sample (training) window, apply them unchanged to the next out-of-sample (testing) window, record the result, then roll both windows forward and repeat.

You judge the strategy on the stitched-together out-of-sample results, the windows the optimization never saw, which approximate what would have happened if you’d re-optimized periodically and traded the parameters forward. It’s a direct defense against overfitting: a strategy that only looks good in-sample, but falls apart out-of-sample, is exposed rather than flattered. The building block underneath it is the parameter sweep, run repeatedly across rolling windows.