Reaction time tests get faster while the athlete gets worse

Repeated cognitive testing produces a learning curve running the opposite way to the fatigue it is meant to catch.

Here is an unhelpful property of every tap-the-screen fatigue test I have watched a squad use: people get better at them. Not fitter, not fresher. Better at the test.

The first week is clumsy. Athletes are learning where the button is, how the timer behaves, how much anticipation they can get away with. By week four they have found the rhythm and their scores have improved by more than any plausible day-to-day fatigue effect. Now stack a genuine fatigue signal on top of that learning curve and watch what happens. Through the hardest block of the season, when the true response should be a decline, the measured response is flat, because practice is pushing up at roughly the rate tiredness is pushing down. The test says everything is fine. The test is cancelling itself out.

This is not a small correction to apply afterwards. Learning curves are steep early and shallow later, so the size of the contamination changes month by month. A squad member who joined in January sits on a completely different part of her curve from someone who has been tapping since July, and their raw scores go in the same column of the same spreadsheet as though they were comparable.

A second effect runs underneath, and it points the other way. Once athletes work out that a slow score buys a lighter session, some of them will produce slow scores. Not conscious cheating, most of the time. Just not trying very hard on a test that carries no reward for trying and a real reward for looking tired. The moment the number affects the training plan, the number stops being an observation of the athlete and becomes part of a negotiation with the coach.

So you have a measure drifting up from practice and drifting down from incentive, and you are hunting a fatigue signal somewhere in between.

What rescues it, partially, is a long familiarisation period that you throw away, individual baselines reset each season, and keeping the score away from whoever decides session load. That last one is the part nobody wants to do, because the whole reason for collecting it was to inform the session.

You can have a clean measurement or an actionable one. Most set-ups quietly choose the second and report it as the first.