Week 1 Flex (RB/WR/TE)

Model run 2026-09-12, Half-PPR. Volume and efficiency from each player's 2026 games to date (nflverse), adjusted for team-mates ruled out on the official injury report, the point spread's effect on run/pass mix, forecast wind, the roof, and home field. Quarterbacks are a 50/50 blend of that model and a gradient-boosted forest trained on three seasons. Backtested on 2024 and 2025: it beats every simple baseline at all four positions. Early in a season it leans on last year, which fades within a game or two. It orders players; it does not predict their exact scores, and nobody's does. VEGAS IS IN THIS LIST: 66 of the 80 players here had a betting line, and for them the projection leans mostly on it. The model-only version is scored against this one every week on the scorecard. MODEL OUTPUT, NOT YET REVIEWED. Generated by the model and published so it can be read and argued with — Reynold has not gone through it line by line and has not overridden anything.