Two tests compared case by case: the verdict, the difference and its range, the cases the challenger fixed and broke, and each check's pass rates.
Passed every time on the champion, not on the challenger.
Per output, called (not reused).
Passed every time on the challenger, not on the champion.
Challenger minus champion, in pass share (−1 to 1).
Cases both runs have.
Two tests compared case by case: the verdict, the difference and its range, the cases the challenger fixed and broke, and each check's pass rates.