The wearable is not the problem, the meeting after it is

Clubs collect better physical data than they have ever had and still make selection calls the same way they did in 1998.

A sports scientist opens a laptop on a Thursday and says a winger's high speed running is up thirty per cent on his four week average and his sprint mechanics have drifted. The manager says he is playing Sunday. Everyone nods. The file is saved. Nobody in that room believed anything different was going to happen, including the scientist, and the reason is structural rather than personal.

Consider who bears the cost of each decision. Rest the player and lose, and the manager is asked in the press conference why his best winger was not on the pitch, and the honest answer, that a number on a screen suggested a raised risk, sounds like an excuse the moment the result goes badly. Play him and he tears a hamstring, and it is bad luck, an injury, one of those things. The incentives are not symmetric and no amount of measurement fixes an asymmetric incentive.

That is why I have stopped finding data quality arguments interesting. Whether a chest unit or an optical system is more accurate at measuring deceleration barely matters if the output feeds into a conversation where one person can override it without giving a reason.

The clubs that get value out of this seem to do one specific thing. They agree the rule before the week starts, in the abstract, when nobody knows which player it will apply to. If a player crosses a threshold, he trains modified on Friday, full stop, agreed in July. Deciding in advance is the whole trick, because the argument on Thursday is never really about physiology, it is about a specific match and a specific opponent and the standings.

There is a real risk of overcorrecting into rule-following, and I would not want a club where a physio's judgement gets outvoted by a threshold either. Data on a single player is noisy, the models come from other squads and other sports, and the honest confidence interval on any individual injury prediction is embarrassingly wide.

But the fix for a wide confidence interval is not to hand the decision to whoever most wants to win on Sunday. Write the policy down, name who can override it, and make the override cost something. Otherwise you have bought a very expensive set of sensors to produce a file that documents your regrets.