AI article
Walk-forward testing with an embargo: keeping time in order
Community description: Random train/test splits leak the future in time series. Here is the walk-forward schedule we use, gap included.
Dev.to | Nov 2, 2026 | Quantooox
Automated excerpt
Training window: the trailing 756 trading days (about three years). Refit: every 63 trading days (about a quarter). So the last training rows have labels that overlap with the first test days.
Selected automatically from source text; not independently written or fact-checked. Read the original for full context.