Screening tests drown in the rarity of the thing they screen for

A test that is right most of the time still produces mostly false alarms when the event it predicts is uncommon.

Take a screening test that correctly identifies eight out of ten athletes who go on to get hurt, and correctly clears eight out of ten who do not. By any ordinary standard that is a good test. Now run it across a squad where five players in fifty will suffer the injury this season.

Four of the five future injuries get flagged. Good. But twenty per cent of the forty-five who stay healthy also get flagged, and that is nine more names. So your alert list has thirteen players on it and four of them are real. Two out of three flagged athletes are being restricted, rehabilitated or worried for nothing, and nobody in the group will ever know which two, because a flagged player who stays healthy looks identical to a correct catch that was successfully prevented.

That last part is the trap that keeps these programmes alive. Every false positive can be reinterpreted as a save. The test flagged him, we intervened, he did not get hurt, therefore the test worked. There is no way to distinguish that story from the true one without a control group nobody is willing to run, and so the programme accumulates evidence for itself out of its own errors.

The arithmetic gets harsher as the injury gets rarer. Screen for something that hits one player in a hundred and even a very accurate test produces alert lists that are almost entirely wrong. This is why the same test can be genuinely useful in a clinic full of symptomatic patients and useless applied to an unselected squad. The test did not change. The base rate did, and the base rate is doing most of the work in the final answer.

So the honest question for any screening battery is not how accurate it is. It is how many people it will flag per real case caught, given how often the injury actually happens here, and whether the intervention that follows a flag is cheap enough to be worth doing nine times unnecessarily.

Sometimes it is. Extra hamstring work for a flagged sprinter costs almost nothing and might help anyway.

Sitting a fit player out of selection because a battery of tests put him in the top quartile of a model is a different bill entirely, and the model has not earned the right to send it.