I hope there's a way to estimate from empirical data how much of the time he's actually right, using a boolean algebra model or something.

Because I'm starting to see a phenomenon, the oppisite action of the decision you make is usually the correct one.