Financial Glossary
Walk-Forward
Testing the model on data it never saw during training. This shows if it works in the real world.
Explanation
Temporal validation: train on N months, test on future M months, advance window. Avoids overfitting.
Technical detail
LearnAImarkets' walk-forward uses an expanding window (train grows with each fold), with an embargo between train end and test start equal to the prediction horizon so labels cannot leak. The public OOS window spans 6 folds.
See also
Educational content. Not investment advice. Past results do not guarantee future returns.