Usage rate hides who actually breaks a defence

The player who ends a possession gets the credit, and the player who caused the collapse gets nothing.

Somebody finishes the play and the box score writes their name down. That is the whole design of the thing, and it is why usage rate is a measure of who shoots rather than a measure of who matters.

Take a possession that starts with the ball going inside. The defence sends a second man. The pass comes out to the wing, swings to the corner, and a shooter who has not touched the ball for twenty seconds takes an open three. In the record of that possession the shooter used it. The passer might get an assist if the ball went directly, and if it swung twice, which is the version that produces the best look, the player who caused all of it appears nowhere at all.

Multiply that across a season and you get a distorted picture of an offence. The number says one player is carrying an enormous load and the others are complementary. Watch the games and you often find the causation runs the other way, that the high usage player is being handed possessions the defence has already been pulled out of shape by, and that removing the cause would remove the effect.

The lineup data catches some of this, which is why on and off numbers are more useful than individual rate stats even though they are noisy and confounded by who else is on the floor. But lineup data cannot tell you the mechanism, and mechanism is what a coach needs in order to decide whether to keep running something.

What I would rather count, and some teams do count internally, is defensive disruption. Possessions where a player's action forced a second defender to commit. Passes made from inside a collapsed defence. Time from the moment the defence broke shape to the moment the shot went up. None of these are on any public site, all of them are countable by a person with a video feed and a notebook, and each of them describes cause rather than conclusion.

Until that exists publicly, the argument about who drives an offence will keep being settled by the statistic that is easiest to collect.

Which is generally how these arguments get settled, and generally why they get settled wrongly.