Walk-Forward Optimization: Anchored vs. Rolling Windows and When Each Fails

I am reorganizing where my quantitative writing lives. Working notes on essays, ideas, and identity stay here on susanpotter.net. Longer-form technical work that maps directly onto the tooling my company builds is moving to referentiallabs.com/blog , where it sits alongside the services and platform pages it informs.

The walk-forward optimization article has moved to Referential Labs: Walk-Forward Optimization .

What the article covers

Walk-forward validation is the closest thing I have to a non-negotiable in my validation pipeline. Every strategy that makes it past exploratory analysis and statistical testing has to survive walk-forward before I take it seriously. The idea is simple; the implementation decisions are where most of the work happens and where most of the mistakes hide. The article covers:

• Anchored versus rolling windows: when each is appropriate, what each implies about the strategy’s edge, and why running both is more informative than picking one.
• The meta-parameter problem: how optimizing training and test window lengths on the results themselves constitutes second-order overfitting, and the discipline of fixing window sizes before running anything.
• Purging and embargo: handling autocorrelation and feature-lookback leakage at training-test boundaries, with concrete purge widths for common signal types.
• Combinatorial purged cross-validation (Lopez de Prado): when it is worth the compute cost and when standard walk-forward is enough.
• What walk-forward can and cannot tell you: passing is evidence, not proof. Its real power is in what it kills.

Why the reorganization

The commercial site is where the tools and services these articles inform actually live. Colocating the writing with the tooling makes the site do a job it should always have been doing, and it lets me use susanpotter.net for the kind of writing that does not belong on a company site: essays, ideas, incidental notes.

Continue to the full article at Referential Labs .