Why keep effect metrics blinded before 300?

Once researchers see the effect estimate, they cannot unsee it. Later choices about thresholds, tolerances, horizons, or cohort definitions can drift toward settings that make the result look better.

300 is not a magic number

In V3R2, reaching 300 fully matched, non-quarantined decisions does not automatically unlock effect metrics. It is only the prerequisite for beginning a manual review.

Why is manual review still required?

Even if the nominal sample count is sufficient, the independent sample count may be smaller due to multiple symbols with the same timestamp, overlapping horizons, or common market shocks. Therefore, dependence, missingness, lineage, and quarantine consistency must be checked even after maturity.

If you want to change the rules

Instead of changing the current arm's threshold after seeing the results, define a new version with a new prospective start. The goal is not to ban better ideas; it is to avoid rewriting the rules of an experiment after the fact.