This is an update to the Net Wins formula.

Thanks for reading NET WINS! Subscribe for free to receive new posts and support my work.

The formula hasn’t changed. The question has.

Net Wins has always been built from one combined number — Player net actions = (PTS + REB + AST + BLK + STL) − (Missed FG + Missed FT + TOV + PF) — allocated to a player by his share of his team’s total, scaled to actual wins. That number has never distinguished between a bucket and a block. A 56-win season built entirely on scoring and a 56-win season built entirely on defense look identical once they land in the same column.

Every term in that formula sorts cleanly into one side of the ball or the other. Points, assists, offensive rebounds, and the misses/turnovers that offense generates all belong on offense. Defensive rebounds, blocks, steals, and the fouls spent trying to get a stop all belong on defense:

Offensive Net = PTS + AST + ORB − Missed FG − Missed FT − TOV Defensive Net = DRB + BLK + STL − PF

Add them together and you get back exactly the same number the site has always used this isn’t a new stat, it’s the old one split down the middle. Which means Offensive Wins and Defensive Wins fall out for free:

Thanks for reading NET WINS! This post is public so feel free to share it.

Offensive Wins = Net Wins × (Offensive Net ÷ Total Net) Defensive Wins = Net Wins × (Defensive Net ÷ Total Net)

As a sanity check, I ran it on the example already in the Net Wins Database: Cade Cunningham, 2025-26, 1,521 net actions, 9.31 Net Wins on a 60-win Detroit team. Split by side of the ball, that’s 7.79 Offensive Wins and 1.52 Defensive Wins a team icon who’s carrying Detroit’s offense far more than its defense, which tracks with how his season actually looked.

Full 2025-26 database (Off Wins / Def Wins for every qualified player) is built. Next up: running the same split back across the full 1946–2026 archive so it’s live in the Explorer.

© 2026 Will Fiore. Net Wins formula and all written content are original works. All rights reserved.