It's an imperfect measure to be sure (peer reviewers are humans), but it's the best criterion anyone has ever come up with so far. It certainly weeds out a lot

It's an imperfect measure to be sure (peer reviewers are humans), but it's the best criterion anyone has ever come up with so far. It certainly weeds out a lot of cranks (even if it on rare occasions mistakes a novel genius for a crank), and isn't that robust to deliberate fraud (peer reviewers generally look for errors in analysis or lack of needed controls rather than look for faked data).