Strokes Gained and the Grey Zones: Reading a Golf Season Through Data

Opening: The Paradox of a Nine-Win Golfer

On the electronic scoreboard at TPC Sawgrass, at midday on March 17, 2026, Scottie Scheffler signed for a final-round 64 to win The Players Championship by one stroke. Three weeks later he pulled on the green jacket at Augusta National. Seven days after that, he won again at the RBC Heritage in Harbour Town. Four months later, at Le Golf National outside Paris, he collected Olympic gold. At the end of the year, he topped the FedExCup.

Nine titles in a single season. That is a number only Tiger Woods at his peak had ever reached.

Yet on the PGA Tour's Strokes Gained tables, in the putting category, Scottie Scheffler's name sits in the bottom half of the list. He wins by moving the ball from tee to green better than anyone else, then compensating for his shortfall on the greens with the very cushion he built up beforehand.

I sat with that data table for a while. What made me stop was not the Scheffler story, but its structure. A golfer can dominate a season while his weakest metric sits in the skill viewers remember most. Spectators remember the putt. The data table remembers the approach.

That is why I am writing this. Not to retell a season, but to reconstruct the way a round of golf is read in a different language.

Context: How Strokes Gained Was Born and Why It Changed the Game

Before 2026, golf statistics on the PGA Tour ran on raw counted metrics: fairways hit, greens in regulation, putts per round. The problem with those metrics is that they cannot distinguish context. A golfer reaching a green in regulation from 210 yards and a golfer reaching one from 120 yards are recorded identically. A putt from one metre and a putt from eight metres both count as one putt.

In 2026, Mark Broadie, a professor at Columbia Business School, published the Strokes Gained method. The core idea is conceptually simple: every position on the course carries an expected number of strokes to hole out. If a golfer moves the ball from a position with an expectation of 3.2 strokes to one with an expectation of 2.5, he has gained 0.7 strokes against the baseline. Sum every shot in a round and you have that round's Strokes Gained.

ShotLink, the PGA Tour's shot-level data-collection system, supplies the raw material. Every shot at every event is logged with coordinates, distance and lie type. From that, Strokes Gained is split into four pillars: Off the Tee, Approach, Around the Green and Putting.

What the method achieves, and what the old metrics could not, is to separate the contribution of each skill. A long hitter who misses fairways will post a high SG: Off the Tee but may be penalised in SG: Approach. A brilliant putter with no power off the tee will look beautiful on the putting tables and fall behind on the overall ones.

Data is never in a hurry; it simply waits for someone who knows how to read it. Strokes Gained took nearly a decade to become a standard, and even now there are analytics rooms in Asia that have not integrated it into their workflow.

Three Decisive Pillars and One Noisy Pillar

Among the four pillars, SG: Approach is the metric most strongly correlated with scoring at the professional level. Research from Data Golf and independent analytics groups over many years points to the same conclusion: the ability to move the ball onto the green from mid and long range decides outcomes more than any other skill.

SG: Off the Tee is the second most influential pillar, but it carries a caveat. Its benefit concentrates in the longest hitters. For the rest of the field, distance does not create that much separation, and accuracy begins to rebalance the equation. This is why a golfer of average driving distance can still lead SG: Off the Tee if he keeps the ball on the fairway.

SG: Around the Green is the least stable pillar from season to season. Studies of the metric's repeatability show a year-to-year correlation markedly lower than SG: Approach. Put another way, short-game skill is heavily influenced by luck and course conditions.

And SG: Putting, the noisiest pillar of all. This is the metric the media loves most and analysts distrust most. The reason is specific: putting is the skill with the highest variance on small samples. A golfer can have one hot putting week and top the table, then three weeks later drop to mid-table with the same technical mechanism.

I have been tracking golf data domestically and regionally since 2026, logging thousands of situations by hand. At one event I followed closely over the first three rounds, a golfer led the putting table at 1.8 strokes per round. Over the next 36 holes that figure turned to minus 0.4. No technical change was recorded. That is variance, not form.

Reading a Season: Scheffler 2026 and Xander Schauffele 2026

Back to Scheffler. Place his four pillars side by side for the 2026 season and the picture is clear. SG: Approach in the tour's leading group. SG: Off the Tee in the leading group. SG: Around the Green at a good level. SG: Putting in the bottom half.

This structure explains something viewers routinely miss. Scheffler does not need to putt well to win, because he generates enough separation at the approach layer that his putts are shorter than his rivals'. A golfer whose average putt is two metres shorter than the rest of the field will have a higher success rate even if his putting skill is average.

This is where the data separates from the feeling. Spectators watch Scheffler and see him miss putts they think he should make. The data table watches Scheffler and sees him standing over easier putts, with a success rate proportional to that difficulty.

People watch the goal; I watch the run before the goal.

Xander Schauffele's 2026 offers the inverse structure, and that is the more interesting part. He won the PGA Championship at Valhalla in May and The Open Championship at Royal Troon in July. Two first majors, at the age of 30.

Looking at Schauffele's Strokes Gained tables that season, no single pillar jumped. What changed was consistency. He had no week with a heavy negative in any pillar. Across four days of a major, consistency is worth more than a peak, because a major is a test of avoiding errors rather than of creating moments.

Before 2026, Schauffele had a long string of major top-10s without a win. Commentators called it a psychological problem. The data table showed something else: he lacked a week in which all four pillars peaked together. In 2026, he had two such weeks.

A report sitting in a drawer is not a conclusion, but a graph waiting for a time axis.

Rory McIlroy and The Masters 2026: When the Time Gap Closes

In April 2026, Rory McIlroy won The Masters at Augusta National in a playoff, completing the career Grand Slam. He had waited eleven years since the 2026 PGA Championship at Valhalla.

The story was retold through emotion: patience, pressure, tears on the final hole. The data table told a different story, in parallel.

Across the decade between his two major titles, McIlroy stayed in the group of golfers with world-leading SG: Off the Tee and SG: Approach. What changed was not skill. It was the structure of Augusta National.

Augusta is the course where distance off the tee creates the largest advantage of any major, because the sloping terrain and fast greens turn short approaches into scoring chances. For years McIlroy hit the ball far enough to contend there but leaked strokes in the Amen Corner holes through accumulated small errors.

In 2026 he did not rebuild his technique. He reduced variance. Final-round data showed no shot landing in the high-risk group at holes 11, 12 and 13, the trio that had cost him chances many times.

That is what the media calls maturity and the analyst calls risk management.

The Tournament System and the Hierarchy Data Does Not Display

To read a golf season you need to know the weight of each event. The professional ladder runs in a fairly clear order: the majors at the top, The Players Championship just below, then the Signature Event group with large purses and limited fields, then regular events, and finally the feeder system including the Korn Ferry Tour.

That weight is not mere prestige. It determines the OWGR points allocated, and OWGR determines major eligibility, exemption status and a golfer's commercial value.

A major win delivers 100 OWGR points to the champion. A win at The Players delivers 80. A Signature Event delivers roughly 60 to 70. A regular event delivers roughly 24 to 40, depending on field quality.

This gap has a consequence viewers do not see: it creates a skewed incentive. A golfer inside the world top 50 can hold his position by playing steadily at regular events, while a more talented golfer who only plays majors will lose ranking through a thin schedule.

This is one of the points where transfer-data models and ranking systems make the same error. Both overvalue young potential and undervalue dressing-room chemistry, or in golf's case, undervalue schedule adaptability. A golfer with beautiful numbers on paper may not hold his ranking if his schedule is not built to optimise points.

The Major Equation: Delivering Under Maximum Pressure

There is a gap in golf data that models handle worst: the gap between regular-event performance and major performance.

Some golfers post comparable Strokes Gained figures in both. Others post good regular-event numbers and collapse at majors. And a small group does the opposite.

Brooks Koepka is the classic example of the third group. At his peak he won four majors in roughly three years, while his regular-event record was far less distinguished. Read only the season Strokes Gained table and he is not the world's number one. Read the Strokes Gained table at majors and he is the most dangerous man in the field.

This raises a methodological problem. If we use season data to predict major results, we will mispredict the Koepka type. If we use major data to predict season results, we will mispredict the far larger group of golfers who play well all year but cannot cross the pressure threshold.

I write the report, close the file, and the market reopens on its own.

My approach is to split the two datasets. One measures baseline ability, using the whole season sample. One measures delivery capacity, using only majors and events of equivalent pressure. When the two disagree, that is a signal worth attention, not an error to discard.

Grey Zones: Correlation Is Not Causation

Here I have to stop at a principle I hold absolutely.

Strokes Gained is a correlational metric. It measures the strength of association between skill and outcome, not causation. When a golfer has high SG: Approach and wins a lot, we may not conclude that raising SG: Approach alone will produce wins. A third variable may be at work, such as training quality, caddie quality, or simply confidence, acting on both.

This error appears frequently in golf analysis. A golfer wins an event leading SG: Putting, and instantly there are articles advising other golfers to practise putting more. But look at the whole season and the share of winners who led SG: Putting is only a small fraction of all winners.

I apply a self-binding rule: a hidden variable must appear at least three times across independent samples before I put it in a report. Three times, not once. This eliminates many interesting findings, and I accept that.

Audiences clap to emotion, but data hears a different rhythm.

The LIV Golf Split and the Problem of Lost Data

In June 2026, LIV Golf launched with a 54-hole format, team play, and financial backing from Saudi Arabia's Public Investment Fund, known as PIF. On June 6, 2026, the PGA Tour and PIF announced a framework agreement to merge commercial interests.

Two years on, negotiations remain unfinished. And in the meantime, a less-discussed problem has emerged: fragmented data.

When a group of golfers moved to LIV, they left the ShotLink system. That means their shot-level data at LIV events is not collected to the same standard. Analysts wanting to compare Jon Rahm's LIV form with his PGA Tour form must work with two datasets of differing resolution.

The clearest consequence sits in OWGR. LIV Golf applied for ranking-point recognition and was rejected, partly because of the 54-hole format and partly because its points-allocation mechanism did not meet the system's criteria.

This is a point I believe the analytics world has underrated. OWGR is not merely a ranking. It is data infrastructure. When an event sits outside that infrastructure, every cross-system comparison becomes unreliable, and a golfer's transfer value is priced against an incomplete set of criteria.

Transfer-data models overvalue young potential and undervalue dressing-room chemistry. In golf, the version of that error is overvaluing season metrics and undervaluing adaptability to a new competitive environment. When a golfer leaves the PGA Tour for LIV, his old Strokes Gained numbers no longer predict much.

Ball Rollback: One Rule That Changes Every Model

There is another factor preparing to change the entire foundation of golf data, and it has nothing to do with commerce.

The USGA and the R&A, the two governing bodies of the global rules of golf, have published a regulation limiting golf-ball flight distance, commonly called the Ball Rollback. It applies at professional events from 2028.

The central idea is to reduce maximum ball flight at high swing speeds, in order to curb the relentless lengthening of golf courses. In terms of impact, models suggest the reduction may run from a few yards to more than ten yards among the fastest swingers, depending on conditions and methodology.

For golf data, this means the entire SG: Off the Tee table loses its value for comparison over time. An average driving distance in 2026 cannot be compared directly with the same figure in 2029.

Analysts will have to rebuild the baseline. And during the transition period there will be a stretch in which the data cannot say anything certain, because the time axis has changed its unit of measure.

This is the kind of risk I call systemic. It does not come from anyone playing badly. It comes from the operating rules of the game changing, and every model built on the old rules must be rewritten.

Slow Play, the Yips and Variables That Never Reach the Table

There are variables that affect outcomes heavily but barely appear in official data.

Slow play is one. Pace of play affects a golfer's rhythm, and rhythm affects outcomes. But no standard metric captures this. A golfer placed in a slow group will wait longer between shots, and for some players that breaks the state of focus.

The yips are the second example. This is the phenomenon of uncontrollable hand movement on short putts, primarily psychological in origin and notably resistant to technical fixes. A former major champion can lose the ability to putt from 1.5 metres for months. The Strokes Gained table records the decline but cannot explain it.

Both cases bring me back to the central principle: data does not replace observation, it supplements it. If I only look at the table, I see a golfer declining in the putting category and recommend a new club. If I also watch the footage, I may see the problem is in the wrist, and the solution is not in the equipment.

An empty stadium does not lack noise; it lacks one dimension of data.

When the Crowd Becomes a Measurable Variable

In 2026, European football restarted in empty stadiums. I was 21, collecting data on 412 matches across five top leagues and comparing it with the five preceding seasons.

The result: home win rate fell from 46 percent to 34 percent, while average goals rose from 2.6 to 3.1. I wrote a 3,000-word piece arguing that the crowd is a measurable 12th player.

That finding was shared by Michael Caley, a well-known analyst, and it opened my first door into the profession.

I bring it up because golf has a similar variable, except nobody has fully quantified it. The crowd at the 16th hole of the Waste Management Phoenix Open generates a level of sonic pressure completely unlike a crowd at a regular event. Tiger Woods once said the noise there helped him focus; younger golfers say the opposite.

This is a real environmental variable. It is not in ShotLink. But it is in the results.

Being pushed out of the game is the fastest way to see the whole board.

The Contrarian Angle: What Everyone Is Reading Wrong

Now the part I consider most important.

There is a widespread belief among golf followers that putting is the decisive skill at the top level. That belief stems from watching: the putt is the shot closest to scoring, and viewers feel its tension more acutely than any other.

The data does not support that belief at the season level.

Across large multi-season samples, SG: Approach correlates with final season position more strongly than SG: Putting. That does not mean putting is unimportant. It means putting explains less variance between golfers over a long season, because its instability causes gains and losses to cancel out.

At the level of a single event, the story reverses. Over four days, a golfer with a peak putting week can win while his approach table is merely decent. That is why weekly forecasting models are far less accurate than seasonal ones.

This is the blind spot of most golf content on social media. It lives in the one-week timeframe, where variance dominates, and it presents variance phenomena as if they were ability.

The second error concerns the evaluation of young golfers. Transfer-data models tend to linearise a young golfer's development trajectory, assuming a rising metric will keep rising. In golf, peak age for most players sits between 28 and 34, and some make late jumps at 30, as Schauffele did.

Undervaluing dressing-room chemistry, or in golf undervaluing the fit between golfer and competitive environment, is the third error. A golfer with beautiful numbers who does not fit the schedule, the course or the support team will not convert metrics into titles.

From Data to Decision: How I Build a Report

A decent golf report, one I write for a team, has four layers.

The first layer is raw data: four-pillar Strokes Gained, average shot distance, putt-distance distribution, scrambling rate. This layer is nothing but numbers.

The second layer is context: course conditions, weather, grass type, course length, and recent schedule. A beautiful SG: Approach figure on a soft course carries a different value from the same figure on a dry, fast one.

The third layer is sample: how many times this variable has appeared, across how many independent samples, and under what conditions. This is where my rule of three applies.

The fourth layer is judgement: a conclusion with an explicit expiry. No "possibly", no "perhaps". Only a verifiable claim and a timeframe in which to verify it.

I do not need recognition in the newsroom; the numbers know their own way to tell the story.

The Match-Report Format and Its Limits

My primary format is the match report: focus on one core finding, reason quickly, conclude clearly.

But I must be blunt about its limits.

Match reports work well when a variable is ripe enough to conclude on. They work poorly when the event is still unfolding and the sample is thin. Pushing a conclusion out early to meet a deadline is the fastest way to turn analysis into an unverifiable prediction.

In golf this is especially dangerous at majors. Major pressure produces results that season data cannot predict, because a golfer's career major sample is often tiny. Four majors a year, and a golfer may play only 20 majors across a whole peak career.

Twenty data points. That is far too small a sample to conclude anything about ability. And that is why I never judge a golfer's major-winning capacity purely on how many majors he has won.

What Comes Next: Signals to Track

From a data perspective, the coming season holds four signals worth watching.

The first is convergence between the PGA Tour and LIV Golf. If the framework talks progress, data infrastructure may be unified, and the entire cross-system comparison problem disappears. If not, golf will continue to exist under two frames of reference, and every global ranking will carry only partial value.

The second is the Ball Rollback. From 2028 the distance baseline shifts, and forecasting models built on earlier data lose comparative validity. That is the moment to rebuild models, not to patch old ones.

The third is the young cohort. If transfer-data models keep overvaluing potential and undervaluing environmental fit, a pricing gap will open. That gap is an opportunity for teams that analyse more carefully.

The fourth is environmental data. Crowds, weather, schedule and course type are variables not yet fully quantified. This is the terrain where I expect many findings in the coming years.

Data is never in a hurry; it simply waits for someone who knows how to read it.

Closing With an Open Question

If you watch a round of golf this weekend and remember only one putt, you are remembering exactly what a spectator should remember.

But if you want to understand why that golfer won, you need to remember the shot that put him in position for that putt. That is the shot the data table cares about. And that is the shot most viewers forget.

The data table does not deny the emotion of a round. It only places that emotion where it belongs inside a larger picture. And in that picture, the only thing that can be verified is what happened, was recorded, and is read correctly.

The numbers do not argue with anyone. They simply wait.