To be fair, I can't access Twitter right now to evaluate what you're saying. But if he's simply saying the model is "wrong" every time the team with >50% chance to win loses then that isn't a very good measure of how good the model is. What you want to see is that when the model says a team has a 90%+ chance of winning it happens 90%+ of the time etc etc.

There are other statistical measures that can be used to evaluate performance. I think the stuff I linked to in the post covers this pretty well.