Surely you mean high-entropy, ie, uniform? We are talking about extremely low-entropy predictions as being the problem here.