matchprior
Free tool

Reliability diagram & calibration checker.

Are your probability forecasts calibrated — does a 70% prediction actually happen about 70% of the time? Paste your predictions and outcomes below and get a reliability diagram, Brier score and calibration error. It runs entirely in your browser — nothing is uploaded.

One row per forecast: predicted_probability, outcome. Probability 0–1 (or 0–100), outcome 0/1 (or loss/win, false/true). A header row is fine. Commas, spaces or tabs.

Reliability diagramn = 0

predicted probability (x) vs. observed frequency (y) · dot size ∝ sample · the closer to the dashed 45° line, the better calibrated

Brier scorelower is better (0–1)
Calibration errormean gap, lower is better
Forecastsrows scored
What you're looking at

Calibration is a number meaning what it says.

Group every forecast by the probability you assigned, then check how often the event actually happened. A well-calibrated forecaster's dots sit on the 45° line: when they say 70%, it happens about 70% of the time. Brier score is the mean squared error of your probabilities; calibration error is the average gap between predicted and observed across bins.

Calibrated ≠ accurate. A forecaster who always says 50% can be perfectly calibrated and tell you nothing. Calibration earns trust in the number; being confident and right is a separate axis.
This is the standard we hold ourselves to. MatchPrior publishes its own reliability diagram, tier by tier, across 67,667 out-of-sample backtested predictions. See our calibration →
Private by design. This tool runs 100% in your browser with plain JavaScript. Your data is never sent anywhere. For information and educational use.