Validation
Walk-Forward Testing
Walk-forward testing repeatedly develops or calibrates a strategy on an earlier window and evaluates it on the next unseen window. The out-of-sample segments are then combined to examine adaptability and stability.
Define the schedule before testing
Choose training length, test length, step size, and parameter-selection rule in advance. Changing the schedule after seeing results adds another layer of selection bias.
Preserve time order
Every training window must precede its test window. Data from the future cannot influence features, parameters, universe selection, or execution assumptions.
Interpret the stitched result carefully
Review each out-of-sample segment as well as the combined curve. Consistency across windows can be more informative than one strong aggregate number.
Research checklist
- Window length
- Retraining frequency
- Parameter-selection rule
- Out-of-sample consistency
Questions and answers
What is walk-forward testing?
It is a sequence of time-ordered training and unseen test windows used to examine how a strategy-development process might adapt through time.