Skip to content
ResetChance
Menu

Data quality

Track record

Which records

Prospective scores use forecasts recorded before the outcome; reconstructed scores are computed on past data. Shadow records come from a model that is not approved.

No records match: 24-hour window, every model, prospective evaluation, published records.

Scores

Not scored yet for every model at once: scores are shown for one model at a time.

Lower scores are better, and only against the baseline on the same forecast origins and target. Forecast frames overlap in time, so they are not independent events; a score here does not show that the forecast is well calibrated. How scores are computed

Reliability groups

Insufficient group data.

Published forecasts

No published forecasts yet.

Every published forecast stays listed, whatever its outcome; later outcomes and revisions are added beside the original number, which never changes.