A two-layer Monte Carlo model built from four SG components: off-the-tee, approach, around-green, and putting, calibrated to every PGA Tour course profile.
Each component addresses a distinct source of predictive information that golf betting markets consistently underweight.
Strokes gained is an accounting identity: total = off-the-tee + approach + around-green + putting. We model each component independently with exponential decay on 4+ years of PGA Tour data, then recombine. No component is lost in the average.
Not all courses favor the same skills. Augusta rewards iron play and precision; TPC Sawgrass rewards driving accuracy. We score every player's historical performance on courses with similar profiles to the week's venue.
We run 100,000+ tournament simulations per event, modeling each player's round-level variance, cut rules, and field strength. Output: true win probability, top-5/10/20 odds, and make-cut probability for the full field.
From raw SG data to a betting edge. Five stages, fully transparent.
We calculate each player's true SG skill level using round-level data from 2018 to present, with an exponential decay half-life of 365 days. Recent rounds count more; rounds from three years ago fade to near-zero weight.
By working at the round level rather than the event level, we capture more data points per player and maintain a better signal-to-noise ratio, especially for players with sparse recent schedules.
We profile every PGA Tour course by which SG components correlate most strongly with performance at that venue. Then we match each player's strengths to the course profile using cosine similarity.
Augusta National heavily rewards approach play. 42% of scoring variance comes from approach. A player who gains 1.5 SG/App but only 0.2 SG/OTT fits Augusta far better than the raw totals suggest.
Recent form matters, but not as much as stable skill. We apply a 50-day half-life momentum component weighted at 33%, blending recent hot or cold streaks with the long-term baseline projection.
A player who has made three top-10 finishes in the last six weeks carries real momentum signal. A player who has missed two straight cuts carries a negative adjustment. The 33% weighting prevents overreaction to small samples.
We run 100,000+ tournament simulations per event. Each simulation draws a round score from a Student-t distribution, which has heavier tails than a normal distribution, matching the true variance of PGA Tour scoring.
Cut rules are applied after round 2. Each player's make-cut probability and post-cut performance are modeled separately, ensuring that make-cut probabilities are not conflated with outright probabilities.
Our probabilities vs. book implied probabilities, devigged using the power method, give us edge on outright markets, head-to-heads, and 3-balls. We size with quarter-Kelly staking to balance edge extraction against variance.
The power devig removes the bookmaker's juice without the inaccuracies of the simpler additive method. What remains is the book's true estimated probability, comparable directly to our model output.
No live betting track record yet, the model launches this season. What we can show is the scope of the underlying data and the architecture that powers it.
No live betting track record yet. Backtested performance is available on request. Live picks begin tournament week. The architecture below reflects the full model as deployed, every component documented, no black boxes.
Pre-tournament outright selections, head-to-heads, and the full Monte Carlo probability sheet, delivered Thursday before first round.
Full pick set released Thursday before the first round begins. Outright win, top-5, top-10, and top-20 plays are all included with model probabilities, edge vs. the market, and recommended unit size.
Beyond outrights, we surface the strongest head-to-head and 3-ball matchups where our probability diverges from the market. These markets offer tighter juice and faster resolution.
Every tournament, subscribers receive the complete probability sheet for every player in the field: win %, top-5/10/20, make cut, and model vs. book edge. You see everything the model sees.
Full-field sheet covers all players with sufficient SG data. Rookies and tour card holders with fewer than 20 qualifying rounds receive regression-to-mean projections flagged with a low-confidence indicator.
All releases land in the Whizard subscriber Discord. The full-field sheet is attached as a formatted table. Individual play cards include model output, edge breakdown, and quarter-Kelly unit recommendation.
Golf picks available weekly with All Access. Every tournament, every market: outrights, head-to-heads, and the full Monte Carlo probability sheet.