Where the numbers come from.

The source

OTIS Golf restates published tournament statistics. Stats, results, and rankings update daily, and more often during an active tournament round.

Coverage and its limits

The stats cover the 15 most recent PGA Tour seasons. Seasons before 2024 ran across two calendar years (2022-2023, for example); the tour switched to single calendar-year seasons starting in 2024. Coverage thins in the older seasons — fewer measured stats were published the further back you go, so early seasons carry results and rankings more completely than Strokes Gained splits.

Strokes Gained stats require shot-tracked rounds. A player won't show Strokes Gained numbers for a season until they have logged enough measured rounds. That's a limitation of the published data, not a gap in OTIS.

Players who have left the tour but played within the 15-season window are kept and marked as former members, so their stats stay visible instead of disappearing.

What we don't do

Every stat on this site traces back to a real published tour number. Nothing is estimated, interpolated, or filled in — a gap stays a visible gap. There is exactly one place where OTIS combines those published numbers into a score of its own, the course-fit model, and it is labelled as a model wherever it appears.

OTIS Golf is an independent project, built and maintained solo, and is not affiliated with or endorsed by the PGA Tour. More about who's behind it on the about page.

Tournament probabilities — and their public report card

The probabilities pageshows win / top-5 / top-10 / make-cut chances for the current field. They come from a disclosed simulation, not a black box: each player's round is drawn from a normal distribution whose center is their adjusted true SGplus a small course-fit tilt (at most ±0.15 strokes per round), and whose spread is their own measured round-to-round variance. Twenty thousand tournaments are simulated in whole strokes, so players finish level as they do on a leaderboard: the 36-hole cut follows the event's own rule and everyone tied on the number plays on, top-5 and top-10 count a tie for the last place, and a tie for first goes to a playoff. The random seed is fixed per event, so a re-run produces identical numbers. The fair odds beside top-5 and top-10 are priced the way a sportsbook settles a tie, by dead heat. Team and match-play weeks get no board and are never graded: they have no finishing order.

Two honesty rules distinguish this from every other model page. First, pre-event snapshots are append-only: once a tournament's probabilities are published they are never rewritten, so the record can't be quietly improved after the fact. Second, the calibration is public: after each event the predictions are graded against what happened — Brier scores and reliability tables, updated on the same page. A probability is a claim about frequency; the track record is the only evidence the claim deserves trust. Players with fewer than 12 measured rounds are priced below tour average and flagged LOW DATA, never silently averaged. These are model outputs, not guarantees.

Adjusted true SG (the skill rating)

The rankings page rates skill with adjusted true SG, built from published per-round strokes gained in three disclosed steps:

• Field adjustment.Per-round strokes gained is measured against that week's field, so +2.0 against a weak field and +2.0 at a major are not the same achievement. Each round is put on an absolute scale by adding the average season SG of the field it was earned against. Rounds whose field strength isn't known stay unadjusted, and each player's “adjusted share” discloses how much of their rating could be adjusted.
• Time decay.Recent rounds count more: a round's weight halves every 180 days, over a two-year window — slower than the form metric, because skill moves slower than form.
• Shrinkage. Thin samples are pulled toward the tour average by their effective sample size, and a player with fewer than 12 measured rounds gets no rating at all rather than a noisy one.

Only measured rounds contribute — non-measured events add nothing, and no strokes-gained value is ever estimated or interpolated. This is a disclosed model in the same spirit as course fit: published inputs, stated weights, honest gaps.

The one thing we model: course fit

The course-fit modelis the single modelled surface on OTIS. It ranks a tournament's field by how well the venue suits each player. Every input is still a published tour stat — the model only decides how to weightthem for this week's course. Every score is a percentile within this week's field (50 is the middle of this field, not an absolute), and it is a fit ranking, not a win probability.

The overall score blends three layers:

• True Skill (45%)— a difficulty-adjusted rating from two seasons of round scores: every round is scored against how hard it actually played, weighted toward the last year (365-day half-life), shrunk toward the field for thin samples, and ranked against this week's field. The same estimator as Form, decayed for skill rather than form.
• Course Fit (35%)— how their game matches this venue: 32% approach, 22% off the tee, 16% around the green, 10% putting, and 20% past results here. Each component is a weighted composite of detailed tour stats (proximity by yardage band, driving precision, scrambling, make rates) chosen by the course's archetype — a distance, balanced, or positional course rewards different skills, and a venue with no hosting history scores neutral on the history part.
• Form (20%)— a difficulty-adjusted strokes-gained measure of recent play: each round scored against how hard it actually played that day, the most recent starts weighted most, and thin samples shrunk toward the field average so a few good rounds can't spike a ranking.

Where a player has little measured data — LIV or major-only players with no recent PGA Tour stats — the missing inputs fall back to below-median defaults (never the field mean), and the row is flagged LOW DATA. Rows whose form rests on only a handful of measured rounds are flagged THIN, and every row shows the rounds of evidence behind its form and its history at the venue, so you can judge how much to trust a given fit rather than taking the number on faith.