At the end in very large cohort standardised testing. And by very large I mean at least hundreds if not thousands of test takers. At that point you have enough

At the end in very large cohort standardised testing. And by very large I mean at least hundreds if not thousands of test takers. At that point you have enough test takers to stack rank them in properly done test. And it smooths things over the years. This is the least bad way of evaluating large number of students. In class room stack ranking is just meaningless. Either student passes demonstrating various levels of knowledge or not.