In general I reject this. Maybe it holds for small classes of statements that regard direct experience (for example "stuff falls down", "getting hit by a car is bad for you") but certainly not for abstract things like theism.

Bear in mind that "more likely" is an extremely weak claim. To say that it certainly doesn't hold for abstract things is, therefore, a rather strong claim. Perhaps stronger than you intend. It would be extremely surprising if humans weren't on average slightly better than random at arriving at correct abstract beliefs even of that type.