Any statistical prove makes certain assumptions and as with every assumption, there are diverging views. In that particular case, the prove ignores any aspects

Any statistical prove makes certain assumptions and as with every assumption, there are diverging views. In that particular case, the prove ignores any aspects that cannot be modelled and then be tested like, e.g., domain knowledge that gives you a headstart over other market participants. As always, I'd say the truth is somewhere in the middle.