Prediction Market Calibration

What calibration measures

Calibration asks a simple question: when markets were priced around X%, how often did they actually resolve YES? Over a large sample, contracts priced at 70¢ should settle YES roughly 70% of the time. Persistent gaps describe systematic pricing behavior — the well-documented favorite-longshot pattern shows up here.

Why sample size matters

A 70% bucket that settled YES 62% of the time is interesting at n=400 and meaningless at n=8. Logit Terminal hides or greys out buckets below a minimum sample size rather than presenting tiny samples as authoritative.

Pinned prices distort calibration

A market sitting at 99¢ minutes before settlement is not a forecast — it is bookkeeping. Calibration built only on final prices overstates accuracy, which is why horizon-based calibration (prices 24h, 4h, and 1h before resolution) is on the roadmap.

How to Read Prediction Market Odds · Methodology · Dashboard