NFL DFS Optimizer and Contest Simulator
Build lineups, then see what actually wins. The optimizer maximises expected points under the real DraftKings rules. The contest simulator does something different and more useful: it fills a tournament with lineups drawn at projected ownership, scores every one of them in thousands of correlated simulated slates, and reports how often a lineup containing each player wins.
Build lineups
Player pool
Click a player to lock him into every lineup, click again to exclude him.
Contest simulator: who actually wins
What the contest simulator does
A projection says how many points a player scores. A tournament does not pay for that. It pays for finishing above everybody else, so what matters is how a player's outcomes line up against what the rest of the field rostered. A twenty five point game from a player half the field owns moves you nowhere. The same twenty five from a player nobody has wins the tournament.
That cannot be read off a projection, off ownership, or off both side by side. So a field of four thousand lineups is built by sampling players at their projected ownership under the real roster and salary rules, every one of those lineups is scored in four thousand correlated simulated slates, and the result is how often a lineup containing each player wins compared with a random entry. A figure of 1.50 means a lineup with him in it wins half again as often as one without.
That single number already contains his projection, his ceiling, his correlation with his team mates and how much of the field is on him. It is what the slate board's leverage column approximates, done properly.
Why the field is sampled rather than optimised
A real tournament is mostly ordinary lineups built around popular players, not sixty thousand copies of the optimal one. Measuring leverage against a field of optimal lineups would answer a question nobody is actually in. The field here also spends nearly all of the salary cap, around 48,800 of 50,000, because an earlier version that sampled purely on ownership left three thousand dollars unspent per lineup and a field that weak makes every entry look likelier to win than it is.
Why correlation matters more here than anywhere else
Lineups are not independent draws. A quarterback and his receiver go off in the same simulated game, so a field full of that stack either wins together or loses together. Scoring a field against uncorrelated player simulations would badly understate how often a stack takes down a tournament, which is exactly the question being asked.
What this inherits
The field is built from projected ownership, and that ownership has no NFL training data behind it yet. Everything on this page inherits that uncertainty. What does not depend on it is the relative shape: who is leveraged against the field as modelled. When real contest results arrive this sharpens without any of it needing to be rewritten.
A player appearing in under one percent of the field has his win rate measured on a handful of lineups rather than on himself, so those are shown but not ranked.