Concordance
Concordance measures how symmetrical a player’s game is, between offense and defense. Do they both score and defend the rim? Would an alternate version of them be the best neutralization? Or is their game imbalanced? Concordance is NOT a measure of quality of contribution, but when cross referenced with contribution it provides context.
Read the formula →“A basketball organism is a single entity, stratified to its characteristics. Both ends of the floor unarguably present distinct habitats, but the Biology Department at the Institute submits that contextualizing a single organism across these habitats is an under-studied feature.”
Foreword
Each context in the graph is a different phase of the play — pick and roll, off-ball motion, post, rim, and iso/self-creation. Each context then gets two ‘spokes,’ one each for offense and defense. Each spoke is weighted so that players’ contributions show relatively similarly across contexts.
An even graph within a context — comparable impact on each end of the court — reads concordant. A spiky pinwheel communicates asymmetric contributions.
- Offense lives at off-ball motion and rim cuts; the other three spokes barely register.
- Defense fans across PnR coverage, post help, and on-ball pressure — the spokes offense doesn't touch.
- The two mostly do not share a context — the pinwheel signature of discordancy.
- Rim is the brightest pair — max on offense, high on defense.
- Defense fans wider into post and off-ball help; offense stays compact.
- Both halves lean toward the same spokes — a smooth roll-and-defend silhouette.
Concordance relies on the best data available. There have been several improvements over recent years, denoted by the number of stars.
Concordance runs from roughly -1 to +1. A positive value means a specimen’s offensive and defensive measured contributions match up in the same contexts, whereas a negative value means they pull in different ones. In other words, concordance, in and of itself, is NOT a measure of quality of contribution, it’s a measure of the symmetry.
The Field Plot · All Recorded Specimens · 1997–2025
Selected Specimens · Plates I–XVIII
Discordant · Thriving
Offense and defense pull at different contexts; positive impact.- Concord.
- -0.23
- RAPTOR
- +1.9
- Concord.
- -0.26
- RAPTOR
- +5.1
- Concord.
- -0.22
- RAPTOR
- +2.2
Neutral · Thriving
Offense and defense draw independent shapes; positive impact.- Concord.
- -0.00
- RAPTOR
- +12.6
- Concord.
- -0.00
- RAPTOR
- +5.1
- Concord.
- +0.02
- RAPTOR
- +3.6
Concordant · Thriving
Offense and defense rhyme; positive impact.- Concord.
- +0.20
- RAPTOR
- +8.6
- Concord.
- +0.16
- RAPTOR
- +1.0
- Concord.
- +0.29
- RAPTOR
- +2.1
Concordant · Faltering
Symmetric silhouette; negative impact.- Concord.
- +0.24
- RAPTOR
- -5.2
- Concord.
- +0.19
- RAPTOR
- -3.8
- Concord.
- +0.19
- RAPTOR
- -3.2
Neutral · Faltering
Independent halves; negative impact.- Concord.
- -0.06
- RAPTOR
- -2.9
- Concord.
- -0.08
- RAPTOR
- -7.6
- Concord.
- -0.06
- RAPTOR
- -2.6
Discordant · Faltering
Asymmetric silhouette; negative impact.- Concord.
- -0.36
- RAPTOR
- -6.4
- Concord.
- -0.41
- RAPTOR
- -0.6
- Concord.
- -0.30
- RAPTOR
- -5.6
Cabinet of Curiosities
Four drawers of specimens worth a second look, set aside for the shapes themselves rather than their station on the field plot above.
Monomania
One spoke towers; the rest stay quiet.One spoke towers; the rest stay quiet.
Rorschach
Offense and defense mirror each other across every context.Offense and defense mirror each other across every context.
Pinwheel
Pure rotational asymmetry — offense peaks where defense is silent.Pure rotational asymmetry — offense peaks where defense is silent.
Pareidolia
Plumages whose silhouettes resemble something else entirely.Plumages whose silhouettes resemble something else entirely.
Catalogue Notes · On the Method
Concordance is computed as the Pearson correlation between the player's offensive action vector and defensive action vector across five contexts — pick & roll, off-ball motion, post play, finishing at the rim, and self-creation — damped by the average absolute gap between the two vectors. Both vectors are scaled 0–1. The damping factor pulls scores toward zero when both halves stay close to baseline, so a flat-everywhere silhouette can't masquerade as a strong concord; a specimen has to be both directionally similar (or opposed) and actively engaged for the score to swing. Values run from −1 (the most opposed fully-engaged silhouette the formula will allow) to +1 (offense and defense fanning around the same spokes at full magnitude).
Effectiveness is measured as RAPTOR where available and HW_RAPTOR otherwise. RAPTOR coverage runs 1977 through 2022; outside that window, HW_RAPTOR — the Institute's ridge-fitted estimate — fills in. The estimate has known calibration gaps for midtier specimens and is itself under audit; readers may treat the Y-axis as approximate for any season after 2022.
The action vectors draw on the existing pre-enrichment tendency suite (Synergy adapters, shot-event action types, BBRef positional minutes) for offense, and on a composite of scheme_versatility, deterrence_effect, help_vs_onball, gambling_tendency — weighted by positions_guarded share — for defense. True matchup tracking is available only from the 2017 tracking era forward; earlier seasons rely on positional proxies.
The plumage glyph rendered alongside each specimen is a 10-axis radial chart with offensive and defensive engagements alternating around the five context spokes. The smoothness of the resulting silhouette is the concordance metric made visible: a smooth fan rhymes; a pinwheeled fan does not.
v1 caveat. The Institute hand-selects a small set of featured specimens whose quadrant is asserted by editorial reading and may not match the rough Pearson formula's classification. The audit toward v2 — calibrating the formula so it surfaces the named specimens organically — is in progress. Specimens outside the featured roster are placed by the formula as recorded.
The qualifying floor is minutes_played ≥ 1500. Below this threshold the contextual mix is too thin for either vector to stabilise; specimens of fewer minutes are noted in field records but excluded from the plate.