Compares scalar confidence with binary correctness, calculates mean squared Brier loss, partitions confidence into caller-selected equal-width bins and reports expected and maximum calibration error. It does not validate labels, correct class imbalance or prove calibration beyond the supplied sample.

Payment Options