The unit is not the possession you know

Almost every basketball rate you have seen is measured per team possession: one trip down the floor, ending in a shot, a turnover, or free throws. Offensive rating, defensive rating, and pace all use it. A good NBA offense scores about 1.1 points per possession on that definition.

Matchup tracking uses something narrower. It records the segment of a possession during which two specific players were matched against each other, and attributes what happened in that segment. If a guard brings the ball up against one defender, gives it up, and the possession ends with a different player scoring against someone else, the first defender's matchup segment contains no points at all.

So the denominator is the same word and a different thing. Across the tracked data, points per matchup possession averages 0.276. That is not a league in a scoring drought. It is a different unit.

Why this ruins comparisons quietly

The failure mode is not that people get confused. It is that the number looks plausible enough to use. A matchup rate of 0.31 sitting next to a team defensive rating of 112 does not announce that the two were measured differently, and nothing about the presentation stops you subtracting one from the other.

The only safe comparison is to another figure in the same unit, and the most useful one is the scorer's own rate against every other defender. That is a within-player comparison: same scorer, same unit, same season, different opponent. It is narrow, and it is the comparison the data can actually support.

How short these segments really are

Matchup rows are small. Across 24,974 tracked rows, a scorer and defender pairing averages 7.24 matchup possessions in a game they both played. That is a handful of trips, spread across a whole night, and it is why single-game matchup lines are close to meaningless on their own.

It gets starker. In 34.2% of tracked rows, the scorer did not attempt a single shot while matched against that defender. A third of the time, the honest summary of a matchup is that it barely happened.

What to do with this

• Read the possession count before the rate, every time. Without it the rate has no meaning.
• Never convert a matchup rate into a per-game points projection. The unit does not support it and the arithmetic will be wrong by a factor you cannot recover.
• Treat any source that publishes matchup rates without a possession count as unusable, not as a shortcut.
• When a matchup rate is compared to something, check what the something is measured in before reading the gap.