Methodology
ARC
ARC units — sd 3.00 per season
ARC = [ 0.65 · z(Impact) + 0.35 · z(Box+) ] · s , s = 3.0 / σ(blend)ARC is Adjusted Regularized Contribution. It estimates the total value a player contributes by blending two measurements of it.
The first is Impact, a ridge-regularized adjusted plus-minus solved over lineup stints, which measures how much the team outscores its opponents while the player is on the floor once the other nine are solved out. The second is Box+, the expected point value of the actions the box score records. The two are not independent: Impact is anchored to a box-score prior at 0.25×, and the halves correlate 0.61, so only 9% of ARC’s spread is box information the lineup solve does not already hold.
Each term is standardised across the qualifying pool so that the two sit on a common scale, combined at a fixed weight, and rescaled so the pool standard deviation is exactly 3.00. A rating of +3.00 is therefore one standard deviation above the pool, and one point of ARC is a third of a deviation. By weight the blend is 0.65 / 0.35; by variance it is 0.68 / 0.32.
Zero is the value of an average possession, not of an average player. Each component is centred on a mean taken over players weighted by the possessions they played, and ARC is the sum of the centred halves, so the offence-and-defence decomposition survives centring without being forced. Because playing time concentrates on better players, the unweighted player mean sits below zero — −1.15 across the 578 players of 2025-26.