# Classifier Reliability

Plots a binary classifier's predicted-probability bins against the observed positive fraction next to a perfect-calibration diagonal, so over- and under-confidence shows as deviation from the reference line; points are drawn `filled`.