The foundation: Strokes Gained
Modern golf analytics is built on Strokes Gained — a statistical framework developed by Columbia professor Mark Broadie that measures a player's performance in each facet of the game relative to the tour average. Every shot is compared to the expected score from that position, accounting for distance, lie, and course conditions.
We track four categories. Together they account for nearly all variance in scoring relative to the field:
Each player's SG profile is computed as a blended rolling average: 65% weight on the last 24 months, 35% weight on the last 12 months. This balances sustained skill with recent form — preventing one bad stretch from tanking a great player's profile while still capturing momentum.
Course-Fit Score (0–100)
A player's raw SG numbers don't tell the whole story. A great driver has more value at a long, open course than at a tight, tree-lined one. A putting specialist is more valuable when the greens are fast and sloped. The Course-Fit Score captures this venue-specific signal.
The calculation has three steps:
1. Weight what the course rewards. Each venue has custom weights for the four SG categories. A course with no preference would be 25% on each.
2. Measure the lean. Each part of a player's game is compared with the rest of the field, then multiplied by how far that category's weight sits from 25%: Fit = Σ (w − 0.25) × z. A player who is equally strong everywhere lands near the middle, however good they are, because Fit measures the shape of a game and not its size.
3. Rank across the field. The result becomes a percentile within this week's field: 100 is the best fit, 0 the weakest. The SG data is blended with a 12-round recent-form window from our own archive.
Course weights: how they're set
The weights are what make this model course-specific. We use a two-step process to set them.
Step 1 — Historical derivation. For each venue, we analyze the last 2–4 years of results and compute the average SG gap between top-10 finishers and the full field in each category. Categories where top finishers consistently outperform the field by a wider margin earn a higher weight. We apply recency weighting (most recent year counts most) and enforce a 10% floor so no category is ignored entirely.
Step 2 — Editorial refinement. Where historical samples are small or the result is counterintuitive, we apply domain knowledge. Firm conditions, weather patterns, and course setup tendencies inform adjustments. The goal is the most accurate representation of what actually wins at each venue.
We currently have custom weight profiles for all major PGA Tour events. Below are a few examples:
| Venue | App | OTT | ARG | Putt |
|---|---|---|---|---|
|
Augusta National
Masters Tournament
|
||||
|
TPC Sawgrass
The Players Championship
|
||||
|
Harbour Town GL
RBC Heritage
|
||||
|
Torrey Pines (South)
Farmers Insurance Open / U.S. Open
|
At Augusta, approach play (32%) carries the most weight — the severe slopes of Amen Corner and the back nine demand precise iron play. At TPC Sawgrass, approach jumps to 42% because the island-green 17th and narrow corridors punish anything less than precision irons. Harbour Town is the most complete shotmaking test on tour, which is why ARG (26%) and putting (22%) are both elevated.
Fit and quality: two numbers, on purpose
Course fit says nothing about how good a player is. So every player gets two separate scores, and we do not add them together.
Fit is the score above: how well the shape of a player's game suits this course.
Quality is how good the player is overall, from blended total strokes gained. It is ranked against every player we rate, about 440 of them, so it means the same thing in a thin field as at a major.
The players worth a look are often the ones where the two disagree: a mid-ranked player whose game suits the course, or a star whose game does not. A single blended number hides both.
Market edge: finding value
Finding a player with a great course fit is only half the job. The bet also needs to be underpriced by the market. We measure this with a simple expected-value calculation:
Edge % = (Model probability × Best decimal odds) − 1
We pull pre-tournament finish probabilities (win, top 5, top 10, top 20) and compare them to the best available odds across DraftKings, FanDuel, BetMGM, Pinnacle, Bet365, and Caesars. A positive edge means the model thinks the market is underestimating this player for this bet type.
Example: if the forecast gives a player a 12% chance to finish top 10, and the best available top-10 odds are +900 (decimal 10.0), the edge is (0.12 × 10.0) − 1 = +20%. That's a bet with positive expected value according to the model.
Data sources
We believe in transparency about where our data comes from.