Confusion matrix
Any system that sorts things into categories (a medical test, a spam filter, a prosthetic hand's decoder) can be audited the same way. Count what really happened against what the system said, and every rate falls out of one table.
The matrix
Rows are what actually happened. Columns are what the system predicted. Click a label to rename it. Correct answers sit on the shaded diagonal.
Per-category rates
Recall (also called sensitivity): of the times it really was this category, how often the system said so. Read along the row.
Precision: of the times the system said this category, how often it was right. Read down the column.
F1: the harmonic mean of the two, a single score that punishes being bad at either.
From a percentage to an experience
A rate only means something once you multiply it by how often the situation happens.