University of San Francisco Chance Details

Calibration of Data-Driven Chances

The Parchment Data-Driven Chances are intended to be well-calibrated. This means that if we tell 100 people that they have a 10% chance, then 10 should be accepted and 90 should be rejected. Or if we tell 100 people that they have a 60% chance, 60 should be accepted and 40 should be rejected.

To use this chart to find out how well calibrated our predictions are for you, identify on the X-axis the range of predictions nearest to yours. Then see what fraction of people with those predictions were actually admitted by looking at the height of the blue (accepted) and green (attending) bars versus the height of the red bar.