2025 NFL WAR: the wins every player adds

Replacement: a replacement-level team would win a quarter of its games. The league therefore holds 0.25 wins per game per team above that level, split 55% among offenses, 40% among defenses and 5% among special teams. Every play is compared with the league average, plus that replacement level. At the end, 35.3 points = 1 win (the regression of wins on point differential, 2020 to 2025).

Offense, in fixed shares chosen by a sensitivity analysis (63 combinations, judged on how well they predict a team's offense the next season as players change teams). A pass to a receiver: quarterback 60%, targeted receiver 25%, offensive line 15%. A dropback without a target (sack, throwaway, scramble): quarterback 75%, line 25%. A run: ball carrier 45%, line 55%. The data barely separate these shares: they remain the most uncertain part of NFL WAR.

Defense, in three layers. Position: the defensive share of replacement value goes to each position by what teams spend on it (edge rushers 25%, cornerbacks 22%, interior linemen 18%, linebackers 18%, safeties 17%), then by snaps. Team: how much less (or more) EPA its defense allowed than average, by snaps. Player: half the EPA of his splash plays (sack, interception, pass defended, forced fumble, tackle for loss, QB hit) and his coverage (yards allowed per target, 0.12 point per yard, measured on play-by-play), each compared with his position's average per snap.

Special teams: 60% to kickers, 40% to punters. Kicker: points over expected by distance (a make-probability curve fitted on every season), plus his extra points against the league rate. Punter: the EPA of his punts against the average.