Season2026·Week1
UpgradeNFL · Week 1
Build a roster, draw thousands of correlated slates, and look at the shape of what comes back. A projection tells you the middle; this tells you the week you win and the week you get blown out.
Full-resolution simulation is an Analyst feature
You can run 2,000 paths on the free tier — enough to see the shape. Analyst raises that to 25,000, which is what you want when the number you care about is a 95th-percentile tail.
Unlock with Analyst — $19/moYour roster
Add
Controls
Press Run to draw 10,000 correlated slates.
Add up nine independent random variables and you get something close to a normal distribution with a narrow spread. That is what naive simulators do, and it is wrong in a specific and expensive way: it makes every lineup look about the same, and it systematically understates both tails.
Football outcomes are not independent. A quarterback and his top receiver share the same passing yards — the correlation is around +0.46. Two running backs in the same backfield split one set of carries — around −0.38. Two players in the same shootout both benefit from the game staying close.
This engine builds the full correlation matrix, factors it with a Cholesky decomposition, and draws jointly. Flip the correlation control to Off and re-run: the median barely moves, but the 95th percentile collapses. That gap is exactly the tournament equity a stack buys you — and the reason people who ignore correlation keep finishing 4,000th.
Marginals are shifted lognormal rather than normal, matched to each player's mean and standard deviation. Fantasy scores cannot go below zero and have a long right tail; a symmetric normal would put real probability mass on impossible outcomes and clip ceilings by roughly 15%.