Movement screens predict almost nothing and cost plenty

Preseason screening has become a ritual that produces a score, a filed report and very little change in who gets hurt.

Every squad goes through it in July. Squats, lunges, hurdle steps, a shoulder reach, each scored on a small scale by a physio with a clipboard, totalled into a single figure. Below a threshold you are flagged. Above it you are cleared.

The tests themselves are reasonable. A person who cannot control a single-leg squat has something worth working on, and finding that out is not a waste of anybody's morning. The claim I object to is the next one, that the total score forecasts who will get injured this season, which is where the whole apparatus gets its budget.

It does not hold up well. The obvious reason is that a screen is a slow, unloaded, unopposed movement performed fresh in a quiet room, and injuries happen fast, fatigued, loaded and with somebody else's body arriving at an unhelpful angle. Whatever the screen measures, it is not the state the athlete is in when the thing goes wrong.

Then there is the composite problem. Adding several sub-scores into one number lets a poor result in one pattern be cancelled by a good result somewhere else, so two athletes with completely different limitations arrive at the same total and get the same recommendation. If the sub-scores mean anything, hiding them inside a sum is perverse.

There is a third issue that nobody likes discussing, which is that these things are scored by humans and the humans disagree with each other. Give two clinicians the same athlete and you will not always get the same number, particularly in the middle of the scale where most people sit and where the flags get raised.

So what would I keep? The conversation. The value of screening day is not the score, it is that somebody watched every player move, asked about last season, found out that a hip has been bothering him since May, and wrote it down. That is a clinical assessment, and it is worth the morning.

What I would drop is the pretence of prediction, along with the ranked list that gets emailed round and quietly influences selection.

The best available predictor of getting injured is having been injured before, and it needs no clipboard at all.