Track prediction confidence against outcomes to improve calibration over time.
This tool requires JavaScript to run.