Chaseit features#chaseit#verification

The Skill page (forecast verification)

At a glance

How often we (and each forecaster) got it right. Rolling skill scores against observed outcomes — kept honest by the case archive.

Deep dive

skill tracks per-forecaster and site-wide hit rates for:

  • Miss rate — no risk forecast, reports observed.
  • False alarm — risk forecast, no reports.
  • Probability of detection — pretty self-explanatory.
  • Brier score — probabilistic calibration over time.

Skill divides by risk category, region, and lead-time. It's what keeps the Outlook honest — if we're routinely over-forecasting tornado in southern England, the score will show it.