Anchored walk-forward vs rolling is a design choice, not a moral one. Searches like rolling window walk-forward and walk forward analysis python implementations often assume one pattern by default. You should pick the pattern that matches how you will research and trade in production.

Rolling walk-forward

Rolling means your in-sample window has fixed length; when time advances, you drop old data from training as you add new data at the end.

Pros: training reflects recent regimes; less contamination from ancient data.

Cons: higher variance if the train window is short.

Anchored (expanding) training

Anchored means training starts at a fixed date and grows over time.

Pros: more data for thin instruments; smoother parameter estimates for slow signals.

Cons: old regimes can dominate; structural breaks can poison the fit.

How to choose

If you will re-fit on a rolling N months in production, mirror that in research.
If your signal is slow and needs long history, anchored may be closer to reality, but add regime checks (Regime change).

Implementation pitfalls

Accidentally mixing anchored IS with inconsistent OOS boundaries
Purge/embargo mistakes when labels overlap (advanced topic; see CPCV vs WFA)

Hybrid designs that work in practice

Some teams use:

anchored IS for slow features that need long history
rolling IS for fast execution filters

If you hybridize, document it explicitly. Otherwise readers assume a clean anchored or rolling story and misinterpret your metrics.

Warm-up and indicator initialization

Rolling windows amplify initialization effects because you repeatedly reset context.

Define warm-up bars consistently and exclude them from objective functions where appropriate.

Write a one-paragraph design memo

Pick anchored or rolling, then justify it with one sentence tied to production refit behavior.

If you cannot justify it, you are likely copying a forum default.