Cut Line

How the numbers work

What's behind the stats on this site, and why they're worked out the way they are. Every number here is measured against the right comparison, shows how much it rests on, and says where it comes from.

Where the numbers come from

Three sources, each for what it's best at, and never mixed in one number or on one chart:

  • Data Golf: each golfer's skill ratings (strokes gained and its parts, driving distance and accuracy), its rankings, the week's field, tee times and the odds. Refreshed through the day; we keep a snapshot each week, so a past week can be judged with what was known then.
  • ESPN: tournament results, round by round, and the money, back to 2001. League results come from here, checked against the old sheets to the cent.
  • Betsperts: strokes gained event by event and course facts. Licensed, so only admins see it for now.

Two sources measure strokes gained against different baselines, so a Data Golf number and a Betsperts number are never added, averaged or drawn on the same chart.

Strokes gained

Strokes gained (SG) is how many shots a round better or worse a golfer is than a baseline. Ours is Data Golf's skill rating: strokes a round against the average PGA Tour player, adjusted for the strength of each field he played (beating a weak field counts for less).

It draws on every round he's played, the latest counting most and rounds two or three years old for very little, so it moves with his form without swinging on one good week. The parts (off the tee, approach, around the green, putting) come from rounds with shot-by-shot data.

+1.50 means about a shot and a half a round better than the average Tour player: six shots over a four-round week.

Percentiles and colours

A number alone doesn't say much: is +0.4 putting good? So most numbers are also placed among every golfer Data Golf rates (about 500, from the PGA Tour, LIV, the DP World Tour and the Korn Ferry Tour). His percentile is the share of them he's better than, a tie counting half.

The colour follows the percentile, one colour for each ten-point band, red at the bottom through orange and yellow to green at the top. A round's strokes gained is placed among that round's field instead.

Because the pool includes the other tours, an exactly average PGA Tour player sits well above the 50th percentile: the average Tour player is better than most rated golfers. So 0.00 strokes gained and a green-ish colour can go together.

The skills radar

The radar shows five parts of a golfer's game on one scale: driving distance, driving accuracy, approach, short game and putting. Yards, percentages and strokes can't sit on one chart, so each spoke is in standard deviations: how far he is from the average PGA Tour player, in units of how spread out all rated golfers are on that skill. It's Data Golf's own scale.

  • The darker ring is the average Tour player (0).
  • The rings outside it are 1.5 and 3 standard deviations better (3 is the rim: very few golfers reach it on any spoke).
  • The ring inside is 1.5 worse, and the centre 3 worse.
  • Each point is coloured by his percentile on that skill, like the numbers beside it.
Distance, +1.8 SDAccuracy, −0.4 SDApproach, +1.2 SDShort game, +0.2 SDPutting, −1.1 SDDistanceAccuracyApproachShort gamePutting
Standard deviations from the average PGA Tour player, 1.5 a ring. The darker ring is average; the rim is three above it. How it works

A made-up golfer: long and a good iron player, a poor putter.

Standard deviations keep the size of an edge: the best putter on tour and the tenth best are both in the top 2%, but the best may be twice as far ahead.

Course fit

Course fit asks: how many strokes a round does this course add to, or take from, a golfer's usual game? A long hitter should gain somewhere that rewards distance more than most courses do. It's worked out in four steps.

  1. What each skill is worth everywhere. We take every rated golfer in every tour event of the last five seasons: his strokes a round against that week's field, and his five radar spokes. Comparing golfers within the same week takes out the weather, the course's difficulty and the field's strength. A regression then says what one standard deviation of each skill is worth, in strokes a round, across the tour.
  2. What each skill is worth at this course. The same, using only this course's weeks. A course has far fewer weeks than the tour, so its answer is noisy; it's pulled toward the tour's answer, harder the fewer weeks it has, so a course with only a few playings stays close to the tour.
  3. Checked on weeks it didn't see. How hard to pull is chosen by hiding one week at a time, working out the course from the rest, and seeing how well it predicts the hidden week. If course-specific numbers don't predict better than the tour's, every fit is zero rather than a guess dressed up as an edge.
  4. His fit. For each skill, how much more (or less) it's worth here than everywhere, times how good he is at it, added up.

When there's a cut, only the rounds everyone played count: the weekend is played by the better half, and measuring against that stronger field would make everyone who played it look worse than they are.

Say distance is worth 0.10 strokes a round more here than on tour and putting 0.05 less. A golfer 2 standard deviations long and 1 below average putting: 0.10 × 2 + (−0.05) × (−1) = +0.25 strokes a round, about a shot over four rounds.

On a course's page, the radar shows what wins there (gold, dashed) beside the average top 10 across the tour (grey, dotted). The gold shape is that same tour top 10, leaning the way this course's top 10s stand out from their own field more (or less) than top 10s usually do. Reading them against their own field matters: a major's top 10s are better at everything because its field is, which says nothing about the course. Each lean is pulled toward none as far as the course's top 10s are too few to tell, so only real differences show. In practice that's distance and approach: the lean on those repeats from one half of a course's years to the other, while accuracy, short game and putting don't, so they stay with the tour. Under it: what each skill is worth here beyond the tour. Lay any golfer over it to compare.

Data Golf's fit first. Our fit can only use finishes, and a course's effect (a tenth of a stroke or so) is hard to tell from the two or three strokes rounds vary anyway. Data Golf builds its course fit from every shot and each hole's length, so when theirs is out for the week, the pick and draft lists, the golfer page and the research table use it instead, and say so.

Its limit: the radars are golfers' ratings today, read against weeks up to five years ago, and today's ratings already include those weeks. That flatters fit a little. We're keeping weekly snapshots of the ratings so it can use each week's ratings at the time.

Average finish and course history

Average finish (a golfer's season, his last five, his record at a tournament or course) is over every start, not just the weekends he played. A missed cut, a withdrawal or a disqualification counts as one place behind everyone in that field: 157th in a field of 156. Leaving them out would make missing cuts look good. The starts and cuts made are shown beside it, so you can see what it's over.

Missed cut, missed cut, T2, in fields of 156: (157 + 157 + 2) ÷ 3 = 105.3, not 2.

Who plays a place well (the world's best on a tournament or course page, and Course horses in the golfers table) is ranked by that average pulled toward a usual level by five starts: (starts × his average here + 5 × usual) ÷ (starts + 5). A golfer with five starts there is judged half on them. In the golfers table the usual level is his own average finish on tour since the start of last season; on a tournament or course page it's the average over every start there by Data Golf's top 50.

Usual level 30th. Two starts at T3: (2 × 3 + 5 × 30) ÷ 7 = 22.3. Five starts averaging 9.2: (5 × 9.2 + 5 × 30) ÷ 10 = 19.6, so the longer record ranks first.

Its limits: a place is a coarse measure (1st to 2nd isn't 40th to 41st), and a field of 70 isn't a field of 156. Strokes gained against the field is the better measure where we have it. The five starts is a convention, not yet checked against held-out weeks.

Odds and the Favorite

Odds are the win odds from FanDuel, through Data Golf, frozen when the first golfer tees off. When FanDuel has no prices for an event, Data Golf's own model odds stand in.

The Favorite, the benchmark every league plays against, takes the week's shortest odds: at a major, where every GM picks two, it takes the two shortest, so beating it is two golfers against two. (The 2026 majors were played before the odds were kept, so the Favorite had one golfer there, against the GMs' two.)

Odds as chances (a setting on your profile) show two numbers, each tagged with whose it is: the book's chance with its margin taken out (FD for FanDuel), and Data Golf's own model's chance (DG). A book's prices add up to more than 100%: that's its margin. We take it out by the power method: every chance is raised to the one power that makes them add up to 100%, which takes more off longshots than favourites, as books shade longshots most. It's worked out when the prices come in and kept with them, for every book we keep (and for prices kept before that, from those prices the same way).

+100, +150 and +300 imply 50%, 40% and 25%: 115%. Raised to the power 1.15 they're 45%, 35% and 20%, adding to 100%. The longshot lost 19% of his chance, the favourite 10%.

A chance shown with a * is a price whose margin we couldn't take out, and runs a little high. When FanDuel has no prices, Data Golf's stand in, shown as Data Golf's; a golfer's odds history keeps FanDuel's only, so one history is one source.

Beat the odds compares where a golfer finished with his expected finish, from Data Golf's chances before the week of each finish (win, top 5, 10, 20 and so on to top 50, and making the cut). It adds up his chance of finishing worse than each place, running in a straight line from his top 50 chance to his make-cut chance at the cut line (the tour's top 65); a missed cut counts as the field + 1, as in average finish. Weeks without Data Golf's chances (before we kept them) don't count. The chalk meter on a GM page is the average win chance (margin out) of their picks: averaging chances, not odds ranks, since 10th in a field of 70 isn't 10th in a field of 156.

Checked out of sample: over 86 PGA Tour events from 2024 on (9,918 golfer-weeks, Data Golf's archived pre-event chances, nothing fitted to them), the expected finish missed by 48.8 places root mean square, against 55.2 for "the Nth favourite should finish Nth", and it's unbiased (golfers finished 0.7 places better than expected on average, against 13.8 places worse than their odds rank). Half of all golfers beat it (50.6%); under the old rule only 11.6% of the week's top 10 favourites ever beat their odds.

Its limits: we keep prices and chances only for golfers we know, so a few of a field's unknowns aren't in the margin we take out (with too few of the field to show a margin, a book's chances are left as they are) or in the field size. Finishes are skewed: a missed cut is far behind, so the expectation sits behind his likeliest finish, and a favourite beats it about two weeks in three. The cut line is the tour's 65 everywhere; a major's differs a little.

If it ended now

While an event is played, a golfer's money "now" is what his place would pay if the event ended this minute: his place's share of the purse by the PGA Tour's standard table (1st 18%, 2nd 10.9%, 3rd 6.9%, and so on down to 65th). Golfers tied share the places they fill equally. A missed cut or a withdrawal pays nothing. It's a snapshot of the moment, not a forecast: nobody expects the leader after one round to win. The majors publish their own tables, so there it's close rather than exact, and an amateur's share isn't passed down.

Each GM's take is then worked out exactly as the real results will be, through the league's rules (a Golden Golfer doubles it, so does a double purse), and the standings with it say where everyone would sit and how far they'd move. The official money replaces all of it as soon as it's in. The projected cut line on Thursday and Friday is the score of the golfer at the event's cut now: the top 65 and ties at most events and the Players, the top 50 at the Masters and the signature events, 60 at the U.S. Open, 70 at the PGA Championship and the Open.

A $10M purse, two tied for 2nd: they share 2nd and 3rd, (10.9% + 6.9%) ÷ 2 = 8.9%, $890,000 each. A GM with one of them as their Golden Golfer would take $1.78M.

Form and process in the league

  • Form in the standings is the average Data Golf strokes gained of each GM's ten best golfers still unused: how strong what's left is.
  • Best process on the stats page is the average Data Golf strokes gained (the number the golfer lists show) of a GM's picks, as it stood the week each was made, so it judges the choice with what was known then, not with how the week went.

GM rates: top 10s, missed cuts, contrarians

A GM's top-10 rate, missed-cut rate and contrarian rate (picks no other GM made that week) are shares of their picks with results. Five picks can say 80% on luck alone, so each rate is pulled toward the league's: we add ten made-up picks at the league's rate before dividing. A full season of picks moves well away from the league; a handful barely does. GMs are ranked by that, and their own count is shown beside it. A GM with fewer than 5 picks isn't ranked. The Contrarian award goes by the same rate, so more picks alone don't win it.

The league's top-10 rate is 50%. Ann has 14 top 10s in 20 picks (70%): (14 + 5) ÷ (20 + 10) = 63%. Bob has 4 in 5 (80%): (4 + 5) ÷ (5 + 10) = 60%. Ann ranks first: her 70% rests on four times the picks.

What we're still fixing

We went through every stat on the site against the rules above. These don't meet them yet, and are being reworked:

  • Best process and pick stats in the draft league don't allow for draft slot: the leader picks last. They'll compare each pick with who was left.

Questions about a number? Ask the commissioner. This page changes whenever a stat does. Back to the leagues.