Calibration: Definition and Formula | CONSENSUS

Calibration

Calibration — the property that events forecast at a given probability occur at that frequency.

By CONSENSUS Research Published Updated

A calibrated model that says 70% a hundred times sees roughly seventy of those events happen. Calibration is checked by bucketing forecasts by stated probability and comparing each bucket against its observed frequency — plotted, this is a reliability diagram.

Why it matters: an uncalibrated model produces probabilities that cannot be used for expected value, because EV depends on the probability being literally true.

Common mistake: assuming calibration implies usefulness. A model that always predicts the base rate is perfectly calibrated and carries no information. Calibration must be read alongside resolution — the ability to separate events that happen from those that do not.

Check any of this against our record

Every signal CONSENSUS publishes carries the bookmaker odds fixed before the event starts and the settled result afterwards — including the drawdowns and the losing runs. The running total is on the front page and every entry is in the log.

See the running result Open the full log