A projection earns its keep in the places it disagrees with last season. Anyone can copy the 2025-26 points table forward; the backtest says doing that misses by about 11 points a player, and the model misses by about 9.5. The difference lives in a few dozen players where the model says last year was not the right number. This post is those players: the forwards whose projected points per game move furthest from their 2025-26 rate, down and up, among those who played at least 41 games and scored 30 points. Per-game, so games missed don't count — this is about what the model thinks of the player, not his health.
The model has three tools, and every row below is some mix of them. It averages three seasons, weighted 6/3/1, so a career year gets pulled toward the two before it. It applies an age curve, which adds up to 16% to a teenager's goals and takes up to 14% off the oldest players'. And it adjusts for the team a player is going to, when he moved. That's it. Nothing here is a scouting opinion.
Trusts less
Every one of these 12 is 33 or older, and that is the age curve doing what it is built to do. Scheifele scored 103 at 32. The model says 87. The year before he scored 87, so the three-season average already sits well under 103, and a 33-year-old then loses 4% to the age curve and most of a minute of ice time (21.1 to 20.3). Malkin and Marchand get the biggest cuts on the board — both over a point a game last season, both projected around 0.77 — because at 40 and 38 the curve is steep and their 2024-25 rates (0.74 and 0.72) were much lower. Kucherov is on this list at 124 and still second overall; 130 points in 76 games is a rate the model does not expect anyone to repeat at 33, but 124 is a long way from a decline.
Which raises the point about the top of the board. McDavid 138 → 129, MacKinnon 127 → 122, Necas 100 → 96 — none of them makes this list, but the model shaves every one. That is regression to the mean, not a forecast of a bad year: a 138-point season is the best case of a 1.5-point-a-game player, and the projection is the average case.
Trusts more
The other direction is mostly down years the model refuses to believe. Michkov had 63 points as a rookie and 51 in year two; the model weights the two seasons, adds the age bump for a 21-year-old (+11% on goals) and 0.60 more minutes, and lands at 65. Peterka (68 then 47) and Monahan (1.06 a game in 2024-25, 0.46 last season) are the same story without the age help. Matthews (1.16 a game in 2024-25, 0.88 last season) is the same story with a bigger name.
Two of the largest moves are trades. Kyrou goes to WSH, and the model's environment factor for that move is 1.13 — the biggest for anyone in this table. Eklund to OTT carries 1.14. The environment number is the model's read of how much a team's shot and goal rates help or hurt a player's own, compared with where he came from; it is the least certain of the three tools and the one to watch when the camp rebuild lands.
And then Cooley and Guenther, both 23 or younger in UTA's top six, both getting the age bump (+11% and +12% on goals) and more ice time (17.3 to 18.0 and 17.2 to 17.9). Guenther at 86 is a top-30 forward on this board after a 73-point season; that is the projection's clearest upward call on a player who did not miss time.
What this is not
It is not a list of who will have a good or bad year. It is a list of where the model's number and last season's number are furthest apart, and why. A 39-year-old can beat his age curve; a 22-year-old can stall. The model's typical miss on a season of points is about ten, and a lot of the rows above are inside that. Ice time is last season's until camp lines are synced, which will move the young players most. The same tables for defensemen are on each player's projection page; the fantasy page shows every player's projection next to his last three seasons.