Rafael Harth comments on How to evaluate (50%) predictions