It is tempting to see high-stakes exams as analogous to a race. Several years of intensive training in preparation for a one-off, winner-takes-all sprint.

Recent reforms reinforce this parallel. The abolition of coursework and modules has drawn a red line between the preparation phase and the test itself. You cover the course, lasting several years, in readiness for the final assessment. Yet so very much depends on how you perform on the day.

At least with athletics, one enters a race knowing its length and layout. With public exams, given interference from the regulator and exam boards, there’s no such certainty. It’s like entering a hurdles race not knowing high how or how widely spaced the hurdles are going to be. Some of the “rules” (e.g., where to draw the grade boundaries) are written after the race has been run.

The recent reforms have added uncertainty, and increased the capacity for surprise. Schools and teachers are expected to track individual progress towards tests over which they have less and less control. During the course, progress is checked against targets and chances graphs based on extrapolations from earlier benchmarks, but even if a student is able and applies herself consistently during the course, things can go wrong after the starting pistol is fired.

‘Meaningless categories’

A fair assessment should hold few surprises. The grade a candidate gets should depend on some combination of ability and application - both measured over a long enough time to be free of random variables. A better athletics analogy would be with a marathon rather than a sprint. The “race” should start when the course begins, not when teaching ends.

One copper-bottomed comparison seemingly remains, that of the candidate in competition with others. However well prepared for the race, there will always be uncertainty about how others perform. The idea that this applies to exams is the final fallacy of high-stakes assessment. GCSEs, even in their reformed state, still rest (albeit precariously) on the principle that if a candidate meets the criteria for a particular grade, then she deserves it, irrespective of the performance of others. Fundamentalists think that this amounts to a bleeding-heart “all shall have prizes” approach, but it lies at the centre of our assessment system, and explains why some element of grade inflation is inevitable.

GCSE reform rested on a flawed idea of what the exam is about. Making the exam “harder”’, and salami-slicing grades at the top end, reflects the desire to rank candidates, as in a race. But given the wafer-thinness of the grade boundaries, and the margin of error involved in marking, it simply cannot be safely assumed that a candidate scoring enough marks to get a 9 is in any meaningful sense more able than another candidate who gets a handful of marks less on the day and ends up with an 8.

The purpose of a public exam at the age of 16 (if such an exam can be justified at all…) should be to identify the broad band in which a candidate falls, to “qualify” them for the next stage, to help sort students into the most appropriate routes, and to inform subject choices. It should seek to “group” candidates by similar levels of ability and attainment, not divide them into ever smaller and more meaningless categories.

Kevin Stannard is director of innovation and learning at the Girls’ Day School Trust. He tweets @KevinStannard1