Matrix SigWhat calibration means
Calibration describes whether stated probabilities match observed frequencies over many comparable forecasts. If a forecaster assigns about 70% to a sufficiently large set of similar events, roughly 70% of those events should occur for that group to look calibrated.
Calibration is not Correct Outcome Rate
Correct Outcome Rate asks how often the most likely outcome was correct. Calibration considers whether the probabilities themselves were appropriate, including forecasts that were right for the wrong level of certainty.
Why sample size matters
One or five forecasts cannot establish a stable calibration pattern. Matrix Sig keeps resolved counts visible and requires 25 resolved predictions for official ranking, while still treating that threshold as eligibility rather than proof of lasting skill.
Extreme probabilities
A 95% forecast makes a stronger claim than a 55% forecast. Repeated extreme probabilities that fail are more informative than a simple win/loss count, which is why probability-aware scoring matters.
Read how scoring works
Review Matrix Sig’s scoring and resolution standard.
Read how scoring works →