Every projection we publish breaks into two numbers you can understand and act on:
Fantasy points = touches × points per touch.
Touches
Touches are a player's chances: targets and carries for running backs, receivers and tight ends; pass attempts and carries for quarterbacks. We forecast them from:
Snap share: how much of the game he's on the field. It's the steadiest week-to-week sign of a player's role.
Touches per snap: how often he gets the ball when he's out there.
Depth chart: for rookies and players on new teams, before they've built a history.
Game script: the betting market's view of how the game will go. Teams protecting a lead run more; teams chasing points throw more.
Points per touch
What a player does with each chance, from his own history, with one important adjustment: we remove touchdown luck. Touchdowns swing wildly from week to week. So instead of rewarding last week's lucky scores, we credit each player with touchdowns at league-average rates for where his touches happen: inside the 10, in the red zone, or elsewhere on the field.
Ranges, not just one number
We simulate every game 10,000 times. That gives each player a floor (a bad week), a ceiling (a great week) and a boom chance (the odds of a week-winning game). Teammates and opponents are simulated together, so when a game turns into a shootout, both sides get the lift.
We grade ourselves
We tested the model on four past seasons it had never seen, and it beat our earlier version. In a blind test on 2026's opening weeks, it matched the major fantasy sites. Every week we publish how our rankings did against the actual results. We'd rather earn your trust than ask for it.
What we don't do
We don't read the news for you. Late injuries and coaching decisions can change everything, so always check the final injury report before kickoff. And we'll never call anything a lock. Fantasy football is gloriously random, and our job is to tilt the odds in your favor, not to promise outcomes.
Scoring: PPR, 6-point touchdowns (including passing). Projections are estimates from a statistical model, not guarantees.