Six-Factor RAPM
A Structural Decomposition of Player Impact
Three independent regressions, six factors, one framework.
The Framework
Standard RAPM gives you one number. It tells you Jokic is a +7.67 — but not why. Six-Factor RAPM opens the box. Take the same ridge regression machinery, the same player-in-lineup design matrix, and run it three separate times — each predicting a different dependent variable: True Shooting %, Turnover Rate, and Offensive Rebound Rate.
Each regression produces offensive and defensive coefficients. Six factors from three models. When you reconstruct full RAPM from these independently estimated components: R² = 0.985. Nearly everything about a player's impact on scoring margin flows through these six channels.
How It Works
Three Independent Regressions
Same design matrix, same stints, same ridge penalty — but three different dependent variables: TS%, TOV%, and OREB%. Each produces offensive and defensive coefficients.
R² = 0.985 Reconstruction
A second-stage regression converts the six factors from native units to points. The result: 98.5% of full RAPM variance explained. The residual is typically < 0.6 per player.
6-Year Time-Decay
Half-life of 700 days. Recent games weighted more, but ~3 seasons of data provide stability. No box-score priors — pure lineup signal. ~1,071 players in the current dataset.
The Six Factors
oTS — Offensive True Shooting
Team shoots better
How much a player raises or lowers their team's shooting efficiency when on court. The single highest-weighted factor — shooting efficiency is the biggest lever in basketball.
What Shows Up
- +Scoring efficiency (volume x efficiency)
- +Spacing and gravity (keeping defenders honest)
- +Playmaking (better passing = better shots for teammates)
- +Ball movement and decision-making
- +Transition pace (easy baskets before defense sets)
What Box Scores Miss
- ?Off-ball gravity that opens driving lanes
- ?Screen-setting quality
- ?Non-spacing centers with negative oTS despite high personal TS%
- ?Hockey assists and secondary creation
Elite Positive
Notable Negative
Cross-Factor Tradeoffs
Non-shooting bigs sacrifice oTS for oREB. Steven Adams: oTS -0.92, oREB +4.04. Joel Embiid: oTS +3.23, oREB -1.45. Completely opposite profiles.
Two Ways to Slice the Grid
Net Efficiency
The quality-of-possessions game. How well do you convert possessions into points, on both ends? Dominated by the TS factors — the biggest lever in basketball.
Curry: elite Net Efficiency, neutral Possession Value.
Possession Control
The quantity-of-possessions game. How many possessions does your team get vs give away? Turnovers and rebounding determine who gets extra chances.
Caruso: strong Possession Value through dTOV, solid efficiency through dTS.
Player Archetypes
Six different mechanical profiles from the 2021-2026 time-decayed dataset — same total impact, completely different routes to get there.
Nikola Jokic
Impact overwhelmingly concentrated in oTS. Modest turnover/rebounding offense. Defense is glass denial — not a rim protector, not a disruptor.
SGA
Positive in five of six factors. Elite in both efficiency (oTS, dTS) and possession control (oTOV, dTOV). The rarest profile.
Rudy Gobert
One dominant factor: dTS. Enormous rim protection with supporting glass work on both ends. Offensive factors are negative or neutral — pure defensive specialist.
Alex Caruso
dTOV is the standout at +3.10 — far and away the highest. dTS is also positive, meaning the gambling isn't costing efficiency. Offense is near-zero across the board.
Stephen Curry
oTS dominance at +4.70 — gravity, spacing, shot-making. But three defensive factors are negative. Net RAPM is 'only' +3.57 because of what the defense gives back.
Draymond Green
Impact spread across multiple factors. dTS +2.34 from communication and scheme, not counting stats. Positive oTS despite modest scoring. Negative oREB from perimeter facilitation.