it would be more precise to say that he argues positions based on evidence which can also support other positions, and therefore isn't convincing evidence to a Bayesian.

What do you mean? Evidence can't support both sides of an argument, so how can one inappropriately use such impossible evidence?