What is walk-forward validation and why is it critical in time series model evaluation?

Walk-forward validation evaluates time series models sequentially, ensuring training only uses past data to predict future points.

Walk-forward validation (also called rolling-origin evaluation) is a time-aware cross-validation method where the training window moves forward in time. At each step, the model is trained on past data and tested on the next time slice. This prevents data leakage and simulates real-world forecasting conditions. It is essential for reliable evaluation because time series data violates i.i.d assumptions.