2025-26 regular season · 82 games · 7,362 FGA. Auto-generated by scouting/build_onepager.py from public data; every number traceable to a gated pipeline in this repo. Shot pricing uses the 2023-24 shot-quality model, whose season-out transfer was validated in the backtest; shot-making is empirical-Bayes shrunk before it is allowed to drive a conclusion.
Terms
PPS: Points per shot: total points on field-goal attempts divided by attempts. Free throws excluded.
xPPS: Expected points per shot - the model's fair price of a shot diet: what an average shooter would score on the same attempts, judged from location and shot type alone, before knowing what went in.
shot-making: Actual minus expected points per 100 shots: conversion above or below what the shot diet itself explains. Positive means making shots the model prices as hard.
EB making: A player's shot-making after empirical-Bayes shrinkage: the observed number pulled toward the league mean in proportion to its sampling noise. The number to trust for projection.
FGA: Field-goal attempts.
RAPM: Regularized adjusted plus-minus: a ridge regression crediting each player with net points per 100 possessions while adjusting for the other nine players on the floor.
net_100: Net points per 100 possessions: 100 x plus-minus / possessions; sits on the same scale as the NBA's published net rating.
rush: Shots in the final four seconds of a period - the desperation window where both shot selection and execution collapse.
empirical Bayes: Estimate the spread of true skill across the league from the data, then pull each individual's noisy estimate toward the league mean in proportion to its noise.
shrinkage weight: The fraction of an observed number that survives empirical-Bayes regression: 1 means fully trusted, 0 means replaced by the league mean.
CI: Confidence interval.
corner three: A three-pointer from the corner, where the NBA line is roughly 3 ft closer than the arc - the geometry discount that makes it the cheapest three.
Keys
They hunt corner threes: 30.2% of their threes come from the corners (league 25.9%). Run shooters off the corner line first.
Mid-range reliant: 13.5% from mid-range vs league 10.0%; live with contested twos over open threes.
Kevin Durant is a genuine shot-maker: +14.8 pts/100 above expectation after empirical-Bayes shrinkage; contest even the bad shots.
Best shot diet: Amen Thompson (1.17 xPPS on 1044 FGA). The value is where the shots come from; deny the spot, not the release.
Their most-used five that wins its minutes: K. Durant - J. Okogie - A. Sengun - J. Smith Jr. - A. Thompson (+9.3/100 over 767 possessions).
Shot profile vs league
HOU
League
Δ
PPS (points per shot)
1.085
1.091
-0.007
xPPS, the shot-diet price*
1.084
1.097
-0.013
Shot-making, pts/100 vs model
+0.1
-0.5
+0.6
Rim share
27.3%
28.4%
-1.1 pp
Mid-range share
13.5%
10.0%
+3.5 pp
Three share
35.0%
41.5%
-6.5 pp
Corner share of threes
30.2%
25.9%
+4.3 pp
Final-4s (rush) shot share
2.0%
1.8%
+0.2 pp
* xPPS prices where they shoot from, independent of whether shots fell: the stable, scoutable part of the profile (year-over-year r ≈ .9).
Personnel (top 8 by volume)
Player
FGA
PPS
xPPS (diet)
Making/100
EB making
RAPM/100
Kevin Durant
1,376
1.176
0.996
+17.9
+14.8
+1.8
Alperen Sengun
1,123
1.074
1.061
+1.2
+0.9
+0.0
Amen Thompson
1,044
1.091
1.172
-8.1
-7.0
+2.3
Jabari Smith Jr.
975
1.080
1.097
-1.7
-1.4
+0.5
Reed Sheppard
946
1.100
1.028
+7.2
+5.1
+1.7
Tari Eason
580
0.991
1.112
-12.1
-8.2
+0.7
Josh Okogie
294
1.061
1.171
-10.9
-5.4
-0.5
Aaron Holiday
252
1.099
1.059
+4.1
+1.3
+1.9
EB making = shot-making regressed to league mean by reliability: the number to trust. RAPM from the opponent-adjusted 2025-26 lineup model (±≈3/100 CI at a season of minutes; treat as descriptive).
Most-used lineups
Lineup
Min
Poss
Net/100
K. Durant - J. Okogie - A. Sengun - J. Smith Jr. - A. Thompson
377
767
+9.3
K. Durant - A. Sengun - J. Smith Jr. - T. Eason - A. Thompson
306
628
+8.8
K. Durant - A. Sengun - J. Smith Jr. - A. Thompson - R. Sheppard