Walk-Forward and Purged Cross-Validation
Once random splits are ruled out, two schemes carry almost all real-world validation of trading research. They answer slightly different questions, and knowing which question each answers is the difference between using them and reciting them.
Walk-forward validation
Walk-forward validation reproduces deployment: train on everything up to a date, test on the period after it, roll forward, repeat.
Each split trains on an expanding (or fixed-length rolling) window and tests on the next slice, so every test observation is predicted by a model that knew only its past. Concatenating the test slices yields an out-of-sample track record, the closest a backtest comes to simulating "what would I actually have experienced."
Its honesty is also its cost. Early folds train on little data and test a model nobody would deploy; only one path through history is evaluated, so the estimate rides on the luck of that ordering; and the scheme consumes data quickly, since each fold's test slice must be long enough to be meaningful. Rolling versus expanding windows adds a genuine modelling choice: rolling adapts to regime change but discards history, expanding does the reverse.
The rest of this lesson is for subscribers
Unlock every lesson in Research Validation and Backtesting, and every other premium course.
Subscribe to continueTest your knowledge
Keep reading Research Validation and Backtesting
21 lessons in this course, and every other premium course, on one subscription.
- Every lesson in every course, with the worked examples and interactive simulators
- Graded questions on every lesson, with explanations for the wrong answers as well as the right one
- The trainers, timed assessments and brainteaser library that go with them