PGA DFS this week
One permanent address for a weekly question. The sheet below is rebuilt for whichever event is on, from the same simulation behind the probability board— so a player’s projected points and their odds of winning are two views of one model.
DFS value sheet
Model outputs and probabilities are information, not betting advice. No projection or probability is a guaranteed winner. 21+. If you or someone you know has a gambling problem, call 1-800-GAMBLER.
Projected fantasy points for the field — expected points, ceiling and floor from the same disclosed simulation behind the probabilities board, scored with DraftKings rules — with an estimated salary, projected ownership and leverage for every player from two further disclosed models. These are model projections, not guarantees, and like the probabilities board this sheet carries its own graded accuracy record below as events complete. Pricing head-to-heads instead? The matchup pricer runs on the same ratings.
How it has scored
Every published sheet is frozen the first time it appears and graded against the field’s actual DraftKings points once the event completes — mean absolute error on expected points, rank correlation on the ordering, and the share of actuals inside each player’s floor–ceiling band. Floor and ceiling are the simulated 10th and 90th percentiles, so about 80% in the band is the design expectation, not an achievement. Nothing is ever back-graded: an event without a frozen sheet is simply absent.
No graded events yet
The record begins with this week’s sheet — graded when the event completes.
What these DFS projections are
Each player in this week's field gets a projected fantasy score under DraftKings golf scoring, along with a realistic ceiling and floor for the same week. The three numbers matter together. An expected score is what the player is worth on average across every way the week could go; the ceiling and floor describe how wide that range is, which is the part a single projection hides. Two players projected identically are not interchangeable if one of them gets there by grinding out top-twenty finishes and the other by alternating wins with missed cuts — and in a tournament format those are very different assets. The projections come from the same simulation behind the win probabilities, so a player's fantasy projection and their odds of winning are two views of one model rather than two models that can disagree.
How the projections are built
The model simulates the tournament many thousands of times from each player's strokes-gained rating and its spread, applies the cut, and scores every simulated week under DraftKings rules — finishing position, the placement bonus, and the streak and hole-score bonuses the format pays. The expected score is the mean across those runs; the ceiling and floor are high and low percentiles of the same distribution, so they are drawn from the simulation rather than invented as a margin around the average. Scoring is set against the course: how many strokes a round easier or harder than an average course it has played in its recent editions moves every player's birdie and bogey rates before skill does. A course with no history is priced as an average one, and the sheet says which case a week is. Salary is not part of the model, because salaries are published by the contest operator and change every week: paste the salary file in and the sheet joins it to the projections and works out value per thousand dollars in your own browser. Nothing pasted is uploaded anywhere.
What the projections do not tell you
They are not lineup advice. Ownership is the single largest factor in a large tournament contest — a high-value player everyone else also rosters is a different proposition from the same player at low ownership — and the sheet does carry a projection of it: signed-in readers see a projected ownership and a leverage figure beside each player, from a disclosed model that follows DraftKings' price first, then win chance and points per dollar. It is a projection of other people's choices, not a count of them, and it is graded only once a settled contest's real ownership is dropped in. The projections also do not price late news: a withdrawal, an injury or a weather delay that reshapes a wave lands after the numbers were generated. Correlation between players is not modelled, so stacking logic is yours to bring. As with everything else the model publishes, the sheet carries its own graded record — how far the projections have missed on average, and how well they ranked the field — as events complete.
Common questions
How are PGA DFS projections calculated?
What is value per $1k in DFS?
Is my salary file uploaded anywhere?
Do these projections account for ownership?
Sourcing is set out on the methodology page; every stat name is defined in the glossary.