Comment by aurareturn

When they run these benchmarks, they should measure false positives - because it's annoying for a real human to solve a captcha.

If blocking 1 bot request is +1 then making a human solve a captcha should be something like -100. That's how we should grade.