The watch is not measuring what the coach thinks it is

Wrist wearables report sleep, recovery and strain with impressive confidence, and most of those figures are inferences stacked on inferences.

A player arrives at training with a recovery score of thirty-four per cent and the staff adjust his morning. The number came off a wrist strap. Nobody in the building could describe how it was calculated, because the company that makes it will not say, and the athlete has quietly reorganised his life around it.

Take the sleep figure first. A wrist device does not observe sleep. It observes movement and, in newer units, some optical estimate of pulse and its variability, and from those it infers whether you were asleep and how deeply. It performs decently at total sleep time in a healthy adult who lies still. It performs poorly at sleep stages, and the deep sleep number that athletes obsess over is the least reliable output on the whole screen.

Heart rate variability is the other pillar, and this one is real physiology being read through an unreliable window. The measure itself is meaningful. The wrist is a bad place to take it, motion corrupts the optical signal, and the value swings depending on when in the night it was sampled, whether you had a drink, the room temperature, and how you were lying. Day-to-day comparisons of one player against himself can carry some information. Comparing two players is close to meaningless.

Then the app takes these inferences, weights them by a private formula, and returns a single percentage with a colour attached. Each step adds error and the presentation removes every trace of it. A confident number without an error bar is the most dangerous object in a performance department.

I am not saying throw them out. Trends over weeks, for one athlete, against his own baseline, do pick up genuine things, and the honest use is as a prompt for a question rather than an answer. A player whose numbers have drifted for ten days is worth talking to.

What is not defensible is a red screen changing a training decision on its own, and it happens constantly, because the number is right there and the alternative is a judgement call somebody has to own.

The strap is a conversation starter that has been promoted to a diagnostic. Demote it.