Expected calibration error: answers grouped by how sure the model was
(10 equal bins), and the average gap between how sure it was and how
often it was right, weighted by bin size. 0 is perfectly calibrated.
Null with no answers that carry a confidence.
Expected calibration error: answers grouped by how sure the model was (10 equal bins), and the average gap between how sure it was and how often it was right, weighted by bin size. 0 is perfectly calibrated. Null with no answers that carry a confidence.