Method
How the measurement works, and what it cannot do.
Every number FogBow produces comes from math a club can check without any proprietary index or any black box involved. This page gives the definitions; a fuller methods note and 4 papers are linked at the bottom.
The statistic
That is the entire calculation. Field scoring average has been computed for decades. This one is deliberately simple, because a measurement a committee cannot verify is a measurement a committee cannot act on.
The three numbers
A golf course is usually described by one number: its rating, or its position in a ranking. Recent work on 3 courses measured every year for decades shows that one number is not enough, and that the missing 2 cannot be inferred from it.
The three numbers. How hard it plays, the level, is where the course sits against its rating. How much it moves is the amplitude. What moves it is the driver profile. Series shown is illustrative.
How hard it plays
the level
The gap between what the course is rated to yield and what it actually yields. Measurable in a single season.
How much it moves
the amplitude
How much that gap shifts from one season to the next. Amplitude is variation between seasons, so a single season contains none of it.
What moves it
the driver profile
The share of the movement attributable to weather, to setup and conditions, and to who played. Different at every course.
Across 15 championship venues measured on the same instrument with fields of the same standard, amplitude ranges from 0.06 strokes to 2.88. That is a spread of nearly 50:1 among the best-run courses in the game. Three courses with nearly identical amplitude — 0.84, 0.79 and 0.85 — can have entirely different drivers. Neither number can be estimated by feel, by reputation, by architect, or by the people who run the course. They have to be measured.
What is held constant, and what is not
Weather
Wind and precipitation come from reanalysis or station records at the course's own coordinates for the specific days played, never a seasonal average. Weather is never folded into the scoring statistic; it is carried alongside it so a reader can see both.
Where a weather coefficient is applied, it is reported as a sensitivity band across plausible values rather than as a single correction, because in championship data the wind coefficient is not distinguishable from zero once venue is held fixed.
Where you measure the weather matters as much as whether you do. Distance to the station is not the test — shared exposure is. At one coastal venue, wind recorded at the official airport station six miles inland explained almost none of the year-to-year variation in scoring, while a shoreline station about a mile from the clubhouse explained half. At an inland venue a station ten miles away gives the same answer as one three miles away. And at a third, the nearer airport reads 2.7 mph calmer than the coastal one, so a wind coefficient measured on the far station has to be rescaled before it means anything. A club that wants to separate what its superintendent did from what the sky did needs a sensor that sees the wind the golfers see.
What is different at a course
Championship fields offer no independent measure of the ability of the players in them, which is why tour analytics has to estimate skill from the scores themselves. An everyday course does not have this problem. Every golfer who posts has a Handicap Index, a calibrated ability measure for every player in the sample, already in the file the course exports.
This applies to resorts and public courses as much as to private clubs, and in one respect more. A club sees broadly the same players year after year; a destination course sees a different field every week, so who played is a far larger share of what moves its scoring average. Destination courses are also disproportionately exposed: coastal, open, and often measured by a weather station miles inland that cannot see the wind governing play. And a visiting golfer has one round to get the course right: if it plays a stroke off its rating, and by a margin that depends on the month, that player is misrated with no chance to recalibrate.
That means each round can be assessed against what that specific golfer, from that specific tee, would be expected to shoot, and the average residual across the membership becomes a difficulty measure that already accounts for who played. It requires no external data, no subscription and no proprietary model.
The known limitation, stated in every report. A Handicap Index is built partly from scores at the player's own course, so a course that gets harder pushes indexes up and partially masks its own effect. Indexes are therefore anchored to a date before the change being measured, or computed from away scores. Which approach was used is always stated.
How much data is needed?
Any golf course has more data than a major championship. A club in a short season may post 9,000 member rounds; a resort or a busy public course can pass 45,000 once visitor play is included. That is 30-150× what a championship produces in a week, and it is why an everyday course can answer questions no championship venue can. Precision improves with the square root of the count, so the first few thousand rounds buy most of it: 1,500 resolve a shift of 0.3 strokes, 9,000 resolve about 0.12, and 45,000 resolve about 0.05.
Why the Course Rating does not answer this
The USGA began rating courses more than a century ago for one reason: to make a handicap mean the same thing at two different clubs. Everything built since has served that purpose. The current obstacle-by-obstacle Course Rating System was designed by Dean Knuth at the Naval Postgraduate School in 1977, almost 50 years ago, and Slope was adopted nationally in 1987. Slope exists because a handicap earned on a gentle course travels badly to a hard one. Portability again.
Nothing in that century of work was built to answer the question a club asks about its own golf course: is it playing differently this year than last, and if so, why. The first mechanism in the American system that considers a single day’s conditions at all is the Playing Conditions Calculation, which arrived in January 2020. It needs at least 8 scores posted that day to run, it is capped between −1 and +3 strokes, and it adjusts the player’s score rather than the course’s rating.
This is not a flaw in the rating system. It is a description of what the rating system was built to do. A Course Rating makes handicaps portable between clubs, and it does that well. It was never meant to tell a green committee whether its own course got harder, by how much, or what moved it. That measurement has simply never existed, which is why the question keeps getting settled by opinion.
Dates from the USGA Course Rating System history and the USGA’s Playing Conditions Calculation documentation.
Where a course stands
Golf’s public rankings mostly measure something other than golf. Exclusivity, prestige, history, pedigree, the clubhouse — all defensible as taste, none of them measurement. Two people can disagree about them forever and neither is wrong, because there is nothing there to be wrong about.
Difficulty is different. It leaves a trace in the scoring record, which means it can be measured, tested, and contradicted.
The grading below comes from the championship studies, which ranked venues on adjusted field scoring average across 4 independent datasets (men’s and women’s Opens, both Amateurs), producing a four-tier ranking that is tested rather than asserted. Read the papers →
At a championship venue the tiers are assigned on agreement across those datasets. A member club has no elite field to be measured against, so the same structure is applied to 3 quantities taken from the club’s own record.
The four tiers
Tier 4 is not an insult. In the championship studies, the Old Course at St Andrews is Tier 4. It is also the most revered golf course on earth. The tier describes the examination a course sets. Not the quality of the golf it offers, not its condition, and not what it is worth to belong to. A Tier 4 course can be, and often is, a wonderful members’ course. What the tier tells a club is what kind of test it is holding, which is a different question and a more useful one.
Why the rating is not the measurement
The same 17 venues, carrying both quantities: what the rating system predicts, and what elite fields actually scored. They do not track each other.
| Venue | Par | Course Rating | Slope | Measured, vs par |
|---|---|---|---|---|
| Lytham | 70 | 76.5 | 152 | +2.06 |
| Merion | 70 | 75.1 | 151 | +4.55 |
| Pinehurst No. 2 | 72 | 77.9 | 149 | +2.94 |
| Birkdale | 70 | 76.5 | 148 | +1.28 |
| Hoylake | 72 | 74.6 | 146 | +0.87 |
| Erin Hills | 72 | 78.1 | 146 | +1.14 |
| Troon | 71 | 76.0 | 145 | +2.38 |
| Shinnecock | 70 | 74.7 | 145 | +3.54 |
| Pebble Beach | 72 | 74.9 | 144 | +2.60 |
| St Andrews | 72 | 75.7 | 143 | −0.11 |
| Oakmont | 70 | 77.6 | 143 | +3.58 |
| Muirfield | 71 | 73.8 | 142 | +3.27 |
| Winged Foot | 72 | 76.4 | 140 | +4.07 |
| Olympic Lake | 71 | 75.8 | 140 | +3.47 |
| Portrush | 72 | 76.2 | 140 | +0.63 |
| Carnoustie | 72 | 75.2 | 139 | +1.35 |
| St George’s | 70 | 75.2 | 138 | +1.89 |
Landmarks rated from championship tees. The 3 highlighted rows are discussed below.
St Andrews and Oakmont are both Slope rated 143. St Andrews played 0.11 under par; Oakmont played 3.58 over. Merion, at 151 the second-highest Slope on the list, produced the hardest scoring of all — and does it under 7,000 yards. A course can be Tier 4 on its posted rating and Tier 1 on measurement. The classification moves because the measurement moves it.
Your tier is yours. FogBow publishes a league table only of clubs that ask to appear in it, and a club can leave it at any time.
Could we do this ourselves?
The arithmetic, yes. It is a mean and a subtraction. Any club with a competent spreadsheet user can compute a field scoring average, and you are welcome to. If your number and ours disagree, we want to know.
The parts that decide whether the number means anything, less easily. Most of the work in an engagement is not the calculation:
What we would ask you not to do is take a number from a spreadsheet into a capital conversation without the second and third of those. The arithmetic is not the risk. An unvalidated number presented with confidence is.
What “adjusted” means, and what it does not
Two of the published papers use the phrase adjusted field scoring average in their titles, and the word has caused confusion, including from AI systems asked to describe the method, which have reported a field-strength normalization model that does not exist. So, plainly:
“Adjusted” describes which rounds are compared with which. It does not describe anything done to the number.
No score is shifted. No coefficient is estimated. No field is normalized against a benchmark. The control is applied at the design stage, by restricting comparisons to events that share a governing body, a qualification route, a format and a defined window, so that the things which would otherwise need adjusting are held constant instead.
In the language of study design this is restriction, not adjustment. Restriction is a legitimate way to control a confounder and it has one advantage over modeling: there is no functional form to get wrong. It also has a cost, which is that it controls only what you successfully hold fixed. Later work uses the word restricted. The published titles cannot be changed.
What this method cannot tell you
What your club would have to send
Published openly, because it is the document that decides whether a course can be measured at all. Nothing here is behind a form or a password.
The published work
4 peer-visible preprints, all with open data.