In the wake of the exams fiasco (“A-level and GCSE students to have downgraded results restored”, August 18), it is worth exploring why the grading system was bound to fail. This was because, entirely reasonably, ministers imposed a constraint on the outcome: that grades should improve by at most 2 per cent. But no constraints were imposed on teachers’ predicted grades.

That was a fundamental mismatch from which only woe could flow. No algorithm can distribute fairly the consequences of an imposed constraint.

How could the system have worked? By distributing the constraint down to the centres and from there to subjects and to teachers. Ofqual would have looked at the historic data for each centre and the prior attainment of cohorts and worked out the expected average grade at each centre. This would be imposed on each centre.

The centre would be required to distribute that average grade across subjects — ie, specify for each subject its average grade so that across all subjects Ofqual’s average was achieved.

Finally each subject would have to allocate grades to candidates so that the average grade was that which the centre had specified.

At each stage there would be a tough negotiation and some room for revision based on evidence. But these negotiations would be among people who knew what was really going on. They would be far better placed to balance the trade-offs among subjects and candidates than any algorithm ever could.

Robert Simons
Eudoxus Systems, Leighton Buzzard
Buckinghamshire, UK