Calibration

Whether stated probabilities match reality: if a model says 70%, do those selections place 70% of the time? Ours is well calibrated across all runners, but not sharp enough on the small set we publish for a per-tip percentage to be worth quoting — which is why we stopped.

← All terms