Should I use a rolling or expanding window for walk-forward? #19
Answered
by
martex-dev
martex-dev
asked this question in
Q&A
|
Should I use a rolling or expanding window for walk-forward? Notes and answer below. |
Answered by
martex-dev
Aug 25, 2026
Replies: 1 comment
|
Expanding if you believe the relationship you are modelling is stable and more data strictly helps. Rolling if you believe the market regime changes and old data is actively misleading. A useful diagnostic is to run both and plot the performance gap over time. If rolling consistently wins, that is evidence of regime dependence and the window length itself becomes a parameter you must validate out of sample. |
0 replies
Answer selected by
martex-dev
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Expanding if you believe the relationship you are modelling is stable and more data strictly helps. Rolling if you believe the market regime changes and old data is actively misleading. A useful diagnostic is to run both and plot the performance gap over time. If rolling consistently wins, that is evidence of regime dependence and the window length itself becomes a parameter you must validate out of sample.