Method
How to research a scorer against a defender
Matchup data answers one narrow question well: how did this scorer produce during the possessions this defender was matched against him, compared with his own rate against everyone else. Reading it correctly takes three things: the sample behind the number, the scorer's own baseline next to it, and the discipline to treat a gap as a reason to watch film rather than as a cause you have established.
What follows is the method Edge applies to its own pages. It publishes no odds, no lines, and no selections, and nothing here predicts a future result.
Start with the unit, or every number will look broken
A possession in matchup data is a partial possession. It counts only the stretch where the two players were actually matched, not a full team trip. A scorer finishes some fraction of the possessions he is guarded on, so points per matchup possession lands far below the roughly 1.1 points per possession of a team offensive rating.
This trips up nearly everyone reading matchup data for the first time, and it has one practical consequence: a matchup rate is only comparable to another rate in the same unit. Never set it beside a per-game average, a team rating, or a figure from a source that has not told you how it counts a possession.
The sample rules, and why they are this strict
A pairing is published here only after 3 meetings and 25 matchup possessions. Between 15 and 25 possessions the lookup tool still answers, labelled a lead rather than a finding. Below 15 it refuses to narrate the number at all and shows the raw totals instead.
Those thresholds are less conservative than they sound. Of 21,324 scorer and defender pairings in the tracked data, 390 clear both bars. Around 98% of the matchups somebody might want to look up do not have enough history behind them to support a claim, and no amount of presentation changes that.
The practical version: if you find a striking split, the first thing to establish is whether it is built on enough possessions to be distinguishable from one good night. Most of the time it is not, and that is the answer rather than a disappointment.
Four questions to ask of any matchup split
How many meetings, and how many possessions?
Two numbers, always read before the rate itself. Three meetings at 25 matchup possessions is the floor at which a pairing gets published here, and it is a floor, not a comfortable sample. A single unusual game moves a 25-possession split hard. If a source shows you a matchup rate without showing you what it rests on, you are being shown a number that could have come from one quarter.
Against what baseline?
A matchup rate alone says nothing. It only carries meaning next to the same scorer's rate against every other defender, measured the same way. A scorer who produces below his own average against one defender is the finding. A scorer who produces below the league average tells you only that he is not the league.
Could the schedule explain it?
Matchup totals absorb everything that happened around them. Rest, back-to-backs, foul trouble, blowouts, injuries on either side, and the coverage the defense was actually playing all sit inside the number, and this data separates none of them. A gap that appears entirely in two blowouts is a fact about blowouts.
Is the defender even the reason?
This is the one most people skip. Matchup tracking assigns the nearest defender, not the scheme. A scorer suppressed while a specific defender was matched on him may have been suppressed by the help behind that defender, by a hard show, or by a coverage the team only plays against that opponent. The data records who was closest, which is not the same as who caused it.
The comparison to avoid entirely
Ranking defenders against each other on points allowed does not work on this data, and it fails in a direction that is easy to miss. Because matchup tracking charges the nearest defender, players who guard the rim absorb the highest-percentage shots in basketball and grade out as the worst defenders in the league.
We measured it: across 338 defenders, the ones such a metric rates worst allowed 50.8% shooting while the ones it rates best allowed 42.9%, climbing without a single reversal. That is shot location wearing a defensive label. Within-scorer comparisons, the kind described on this page, do not have the problem.
The full analysis, with the method to check any leaderboardWhat this data cannot do
- It does not predict. Every figure describes possessions already played, and past performance does not establish what happens next.
- It does not isolate cause. A gap between a matchup rate and a baseline is an association, and scheme, help, rest, and shot variance all sit inside it unseparated.
- It does not know tonight's context. Injuries, rotations, minutes, and coverage plans are not in it, and those often matter more than the history.
- It does not carry a price. Nothing here is compared against a line, because no line is published or linked anywhere in this section.
Used well, a matchup split narrows what to look at. It tells you which possessions are worth pulling up and which questions are worth asking of the film. Treated as an answer on its own, it is a number with a lot of unexamined things inside it.
Apply it
Run the method on real pairings
Every published pairing, with the meetings and possessions on each row, sortable by the gap between a scorer's rate against a defender and his own baseline.