Set piece models are running on far too few corners to say much

A season gives a club perhaps two hundred corners, which is nowhere near enough to grade a marking scheme.

Two hundred corners, spread over a season, split across two marking schemes and half a dozen personnel changes.

Try to get a reliable estimate of anything out of that. You cannot, and yet set piece defending is graded confidently every week, usually on goals conceded from corners, which is an even smaller number. A club might concede four or five in a league season. Four events. People write about the collapse of a defensive system on the strength of four events.

Expected goals from set pieces helps, because it uses every attempt rather than only the ones that went in, and it lifts the effective sample from a handful to something in the dozens. Still small. Set piece xG models also have their own difficulty: the shot quality distribution from a crowded six yard box is unusual, deflections and blocks are common, and the features that ordinarily predict conversion behave differently when eleven bodies are packed into a small area. Models fitted mostly on open play are being asked to extrapolate.

The honest position is that a single season cannot separate a well organised zonal scheme from a poorly organised one at the level of outcomes. Two or three seasons might, if the scheme and the personnel stayed constant, which they never do.

That leaves process measures, and this is where I would put the effort. Did the marking assignments hold. Was the first contact won. How far from goal did the ball get flicked. Where did second balls land relative to the players positioned for them. Those things happen on every corner, not just on the four that ended in a goal, so the sample is two hundred rather than four, and they can be measured from tracking without waiting for outcomes to accumulate.

First contact rate alone would tell you more about a marking scheme than a season of goals conceded, and it would tell you sooner.

There is a broader point about how football discusses set pieces. The events are rare, the outcomes are dramatic, and the human tendency is to explain rare dramatic outcomes with structural causes. Sometimes the structure was wrong. Sometimes a defender slipped.

With four data points you have no way to tell the difference, and the confident weekly verdicts are mostly noise being read aloud.