Gym numbers and the coaching eye

How the profession divided itself

Somewhere in the last two decades strength and conditioning settled into a comfortable division of labour. The numbers are the science: force plates, timing gates, velocity-based training, GPS, profiles and dashboards. The watching is the art: the old coach at the side of the track who says the athlete looks heavy today and turns out to be right.

Art, in that sentence, is doing a lot of quiet work. It sounds respectful. What it actually means is unaccountable — nice if you have it, impossible to teach, and not really evidence. The effect of the label is that one half of the job gets budgets, courses, journals and job titles, and the other half gets described as a gift.

I want to argue that this is backwards in an important way. Not that the numbers are worthless — I am doing a master's degree in them and I use them every week. But the confidence we place in them is often less well founded than the confidence we place in trained perception, and we have organised our profession as though the reverse were true.

We measure what is easy to measure

Start with the obvious thing that nobody says out loud. We do not measure what matters most. We measure what a device happens to be able to capture.

Nobody chose the countermovement jump because it is the most important quality in sprinting. It became ubiquitous because force plates measure it cleanly, it takes twenty seconds, and it produces a number that goes up and down. The same is true of most of the standard battery. The measurement set in our field is a history of available technology, not a considered account of what produces speed.

Choosing what to measure is an opinion about what matters. It just arrives dressed as neutrality.

And the numbers themselves are frequently softer than they look. In writing about sled loading I went through the reliability work on sprint force-velocity profiling, where the two variables coaches most want to act on — the slope of the profile and the rate of decrease in force ratio — have been shown to vary so much between sessions that researchers questioned their use for individual prescription. In the piece on high-speed running, the metric that ran professional sport for a decade was eventually dismantled by its own statistics.

These are not fringe examples. They are two of the most influential numbers of the last ten years, and both turned out to carry far less information than the confidence around them implied. Meanwhile the coach who said the athlete looked flat was reporting a real observation.

The number as a shield

Here is the uncomfortable part, and I include myself in it.

A number does not only tell you something. It also protects you. If I pull a player out of a session because the data flagged them, and they get injured anyway, I followed the process. If I pull them because they looked wrong to me, and I am wrong, I look like a coach guessing with someone's career. In an environment where practitioners get moved on quickly, the incentive to defer to the dashboard is not really about the athlete at all.

That is worth naming honestly, because it explains something the evidence cannot. It explains why metrics survive long after their limitations are published. Their function in the room was never purely informational. They settle arguments, distribute blame, and make a decision look defensible to people who were not at the session. None of that requires the number to be true.

The eye is trained perception, not mysticism

The other half of the argument is that we badly misdescribe what the experienced coach is doing.

When a coach watches a rep and says the athlete is landing slightly ahead of themselves, that is not intuition in any spooky sense. It is pattern recognition built from tens of thousands of previous reps, operating faster than language. It is the same faculty a radiologist uses to spot something on a scan before they can articulate why, or a musician uses to hear that a note is flat. In every one of those fields we call it expertise. In coaching we call it art and treat it as decoration.

It has a further advantage the dashboard does not. Perception is continuous and contextual. The force plate gives you one number from one morning. The coach's eye is running for the whole session, taking in gait, mood, how the athlete walked in, whether they are talking less than usual, and what they did to the last rep of the last set. No monitoring system available to a private group covers a fraction of that.

Nobody trains the thing we call art

So here is my actual complaint. If the eye is a perceptual skill, it is trainable — and almost nobody trains it deliberately.

Perceptual expertise develops under specific conditions: many repetitions, prompt feedback, and honest records of when you were wrong. Radiologists get that from biopsy results. Coaches get almost none of it. We watch thousands of reps and receive feedback on approximately none, which is why coaches can accumulate twenty years of experience and not improve after the fifth.

The fix is not complicated and I have started doing it. Before the gates or the app give me a number, I write down my call. That rep looked slower. That athlete is 2% down today. He has stopped extending on the left. Then I look at the data and mark myself. Over a season this does two things: it tells me which of my perceptions are reliable and which are stories I tell myself, and it makes the eye auditable — the exact property the numbers were supposed to have and the art was assumed to lack.

It also changes the relationship between the two. The data stops being the verdict and becomes the feedback loop that sharpens the perception. That is a far better use for it than telling me what I already saw.

A test for your own practice

Two questions. First: when the number and the athlete in front of you disagree, which one do you actually believe? Not which one you would defend in a meeting — which one changes what you do next. If the honest answer is always the number, you are not coaching, you are administering.

Second: would you still collect this metric if it took an hour of session time instead of thirty seconds? Most of what we measure is collected because it is nearly free, not because it earns its place. Anything that fails that test is habit dressed as rigour.

My Level 2 taught me to see. The MSc taught me to doubt what I see. The mistake our profession keeps making is assuming the second one replaces the first, when the entire value of the doubt is that it makes the seeing better.