Aggregate predictions by confidence level and compare stated

Aggregate predictions by confidence level and compare stated probability to actual frequency across dozens of cases rather than evaluating single predictions, as systematic patterns reveal calibration while individual outcomes contain too much noise.