What is walk-forward testing and when should I use it?

Question

AcadiFi · Accepted Answer

Walk-forward analysis (rolling-window backtesting) re-estimates the model periodically as new data becomes available, mimicking how a practitioner would actually deploy the strategy. Structure: (1) train on periods 1-N; (2) test on period N+1; (3) roll forward — train on 2-(N+1), test on N+2; (4) continue. You get a concatenated out-of-sample return series. Two variants: expanding window (training window grows) and rolling window (fixed-length training). Example: Caldera Capital tests a sector-rotation strategy on 2005-2025 monthly data. Rolling 60-month training window + 12-month test, stepping annually: train 2005-2009 → test 2010; train 2006-2010 → test 2011; and so on through 2025. Concatenated OOS Sharpe = 0.8 vs in-sample 1.4 — realistic degradation. Advantages over single train/test: (1) tests robustness across many market regimes; (2) reveals whether parameters need to drift over time; (3) produces a tradable-looking return series. Caveats: (1) you still have hyperparameter choices that can be mined; (2) window length itself is a parameter; (3) too-frequent retraining risks overfitting to recent noise. Walk-forward is standard for systematic strategy deployment at funds like AQR and Two Sigma.

What is walk-forward testing and when should I use it?

Master Level II with our CFA Course

Related Questions

Practice Questions