Most NBA matchups have no usable history
87% of NBA scorer and defender pairings have under 15 matchup possessions, and the median pairing is a single game. Here is the full distribution.
5 min readResearch
What the matchup data can and cannot carry
Measured analyses of NBA scorer versus defender tracking, written for people who want to know how far a number can be pushed before it stops meaning anything. Every figure was computed against the matchup table, and every article states the method beside the claim.
Several of these are negative results. They are here because a measure that does not work is more useful to know about than one more leaderboard, and because the alternative is publishing numbers we cannot defend.
Start here
What a matchup possession actually measures
NBA matchup possessions are partial possessions, counting only the stretch two players were matched. That is why the rates look far lower than a team offensive rating.
Read the analysis
Every extreme in matchup data is a small sample
Variation in NBA matchup rates falls fourfold as the sample grows, and 55% of thin pairings are scoreless. The eye-catching numbers only exist where the evidence is weakest.
6 min read
Do NBA matchup splits repeat? A split-half test
Splitting each NBA pairing's games in half and correlating the two halves gives r = 0.44. Corrected for length, reliability is about 0.61: real signal, and roughly 40% that will not repeat.
7 min read
Which NBA scorers look most matchup-dependent
Jalen Brunson's rate ranges from 0.065 to 0.842 across nine defenders. Here is the spread for the league's most variable scorers, and how much of it is noise.
6 min read
What NBA matchup tracking actually records
NBA matchup tracking assigns the nearest defender, not the scheme. It records proximity and outcome, and leaves out help, coverage, and intent, which is where most misreadings begin.
6 min readHow to read anything here
Matchup rates are measured in partial possessions, counting only the stretch two players were matched, so they run far below a team offensive rating and are not comparable to one. Every rate is only meaningful next to the same scorer's own rate against other defenders.
Underlying data was last refreshed 7 June 2026 and the analyses were computed 27 July 2026. Nothing here forecasts a future result, and no odds, lines, or selections are published anywhere in this section.