How team defence is measured

How is team defence measured on this site?

In three pillars, over 180 team-seasons, each ordered by its own measured reliability rather than by how interesting it reads. This is a team metric on purpose: a shot split five ways across defenders is team defence wearing a player's name.

The three pillars, ranked by how much they can be trusted

Split-half reliability of each defensive pillar, 180 team-seasons
PillarUnitReliability
Quality forcedexpected points per shot faced, vs league mean0.981
Rim deterrencepercentage points of shots faced at the rim, vs league0.964
Conversion suppressionpoints below expectation per 100 defensive possessions0.717

Each figure travels with its pillar so the noisiest one cannot be read as the solid one.

Why there is no player version

There was one, and it was rejected on evidence: it split-halved at 0.348 against 0.981 for the team version. Shots split five ways across defenders makes the player version a team metric with shared credit; its leaderboard placed non-defenders near the top. D-RAPM in Layer 1 is the regression-adjusted player defensive signal.

League-centred, and what is left out

Quality conceded and rim rate are expressed against the same season's league mean, so pace and rule changes do not leak into a team's figure.

This layer needs shot-level expected-points against a reconstructed five-man defensive lineup. No European league publishes the possession-level data it requires, so there is nothing here to approximate from. Absent, not pending.

Over Expected is a basketball shot-value platform covering NBA, EuroLeague, EuroCup, Liga ACB, Lega Basket Serie A, Basketball Bundesliga, Greek Basket League, ABA Liga, WNBA.