The fraction of test cases where you can definitively determine if the output is correct, as opposed to cases where correctness is unknown.