The theory is ROC curves [1]. The practice requires that you play devil's advocate with yourself - figuring out ways to disconfirm your own beliefs.

I usually start with the question: 'What data would convince me that I am wrong?' If I can't answer that for myself - I know I have a problem.

[1] https://en.wikipedia.org/wiki/Receiver_operating_characteris...