The ShotLink Gap and Professional Golf's Empty-Data Problem

2 a.m. in Nagoya, and I reopen the ShotLink board from the final round of a PGA Tour event that has just ended. Fourteen statistical columns line up in perfect order: average driving distance, fairway rate, GIR rate, putts per round, and Strokes Gained broken down by area. I need exactly one thing that is not there: how many strokes the leader lost across the last six holes, when fitness and pressure tighten at the same time. The board has no such column, and it never has.

I am writing this to talk about something professional golf analysis rarely admits: the biggest decisions in this sport — from the new ball rule to major eligibility — are being made on data tables with holes in them, and those holes are usually left unmarked. An empty column in a data table is not automatically a confession. It is simply empty. The reader of numbers has to notice it alone.

Context: the decade golf learned to count

The data revolution in golf has a fairly clear starting point. In 2026, the PGA Tour put ShotLink into operation — a system that records every shot, every ball position, every remaining distance to the pin. Before that, golf statistics were crude: stroke totals, fairways hit, putts. ShotLink turned each round into thousands of geolocated data points and, for the first time, allowed the question to be not "how many strokes did this player take" but "was this shot good or bad against the baseline".

Then came Mark Broadie, a Columbia Business School professor who introduced the Strokes Gained metric and systematized it in his 2026 book "Every Shot Counts". The core idea is simple and ruthless: a shot is measured not by its outcome but by the difference between that outcome and the tour average from the same distance and lie. A three-metre putt holed is not praiseworthy if the tour average also holes it. An eight-metre putt missed is not blameworthy if the tour average also misses it.

Strokes Gained splits the game into four areas: Off the Tee, Approach, Around the Green, Putting. From there, one can say precisely where a player is strong and where weak. That is a genuine advance. It is also a genuine simplification, and I will return to that point later.

After ShotLink and Strokes Gained came Data Golf, an independent analytics platform built by brothers Matt and Will Courchene, offering alternative SG models and predictive ratings. Then came countless in-house models from teams, bookmakers, and investment funds. Golf, long seen as the most conservative and mystical of sports, suddenly became one of the most heavily measured on the planet.

That is the success story. The rest of this piece is the story of its cost.

ShotLink does not cover everywhere

ShotLink is the PGA Tour's system, and it mainly covers PGA Tour events. Many DP World Tour events, most regional-tour events, and almost all of women's and elite amateur golf have no shot-by-shot data. Which means that when someone says "player X's Strokes Gained is poor", they are speaking about a far smaller sample than the reader imagines.

In Japan, where I live and work, the JGTO — the country's professional men's tour — has better data than most Asian tours, but it is still not identical to ShotLink. When I try to map a player's JGTO metrics onto a PGA Tour SG model, I have to accept an error that can reach a few tenths of a stroke per round. A few tenths of a stroke per round, times 72 holes, is the distance between first and thirtieth on the leaderboard.

That gap is not ShotLink's fault. It is the nature of a system designed for one tour and then taken global for comparison. This is the first time I have to remind myself of the line: The gap in the data table can speak too, if we are willing to listen. What the gap says here is that any cross-tour comparison must carry a note on data source and error margin.

It is also worth stating plainly what the board displays but few read closely. GIR — greens in regulation — counts holes where the ball reaches the green within the allotted strokes minus two. But it does not distinguish an iron shot that finishes two metres from the pin from a wood shot that finishes twenty metres away. Both are GIR. Both count the same. Yet distance to the pin is what governs the putt count. The board counts the easy thing and ignores the harder thing that matters more.

OWGR: a ranking system with a blocked horizon

The Official World Golf Ranking (OWGR) was created in 2026 and is the basis for major-championship and elite-event eligibility. Its points depend on a player's finishing position at each event and the quality of that event, with weighting that decays over time.

The problem arose when LIV Golf, backed by Saudi Arabia's Public Investment Fund (PIF), launched in 2026 and pulled a string of top players away from the PGA Tour. LIV events were initially denied OWGR points, meaning a former world number one could slide down the ranking while still playing at an elite level. The consequence is that major eligibility — tied to OWGR — became a political question more than a technical one.

On June 6, 2026, the PGA Tour and PIF announced a shock framework agreement, declaring the two sides would merge their commercial operations. Since then the process has remained unfinished and volatile. But the point I want to keep is not the timeline; it is how data is used as a weapon: whoever controls the ranking system controls the horizon of a player's career. A player can produce the best golf of his life inside a system that awards no points, and on paper, he is in decline.

This is where I apply a lesson from the past. In 2026, while working as a data contributor for a football site in Nagoya, I analysed the Japan–Belgium round-of-16 World Cup match using PPDA and concluded Japan pressed well. I ignored Belgium's running distance after the 70th minute. Belgium came back to win 3-2. I was wrong, and I admitted it publicly. The lesson is not about football. It is that a metric system can be highly detailed and still miss the exact decisive variable. OWGR is such a system, at a global scale.

Ball Rollback: when past data argues against the future

On December 6, 2026, the USGA and The R&A announced a new ball rule limiting ball flight distance under test conditions. It applies to elite competition from January 2028 and to recreational play from January 2030.

The argument behind it rests on decades of steadily rising driving-distance data. But here the data table has a large hole: the strongest distance data belongs to professional players, while the rule's impact spreads across tens of millions of recreational golfers, a group for which detailed data barely exists. A decision for the entire sport is being made on data from a very thin layer at the top.

I do not oppose the Ball Rollback. I oppose how it is presented as an unassailable data conclusion. Data is never wrong; it is I who asked the wrong question. The right question is not "how far does the ball fly", but "how far does the ball fly, for whom, at what level, and how does that change the experience of play". The data table answers the first question very well and the second very poorly. The distance between those two answers is where the entire ball-rule debate lives, and it will run at least until 2028.

Strokes Gained forgets variance

Back to Strokes Gained. It is the metric I use daily and also the one I distrust most. It is an average. A player with a positive SG: Putting for the season may mean he putts steadily, or it may mean he had three explosive putting rounds and twenty bad ones. Those two stories lead to completely different predictions for the next round, but Strokes Gained shows the same number.

In golf this is especially dangerous in Putting, where the sample is small and the variance largest. A streak of holed putts across two rounds is not skill; it is noise. But if you look only at the SG: Putting column, you will not see noise. You will see a player "finding form". And you will back him next round on a sample too small to say anything.

I once made exactly this error in a hand-built xG model back in 2026 for a football club playing in Japan's second division. I omitted the home-ground factor in a four-match losing streak and got 6 of the last 10 rounds wrong. I had to review all the footage, check every phase of play, and accept that raw data was not enough. I began writing by reverse verification from then on: never give a number without a contextual condition.

Golf needs exactly that method. Every SG metric should come with standard deviation, sample size, and time window. Every number is a confession not yet written into prose — and the analyst's job is to write that confession out, not to leave it sitting silently in a spreadsheet cell.

There is one more layer Strokes Gained does not touch: the FedExCup. This is the PGA Tour's season-long points system, launched in 2026, ending with the Playoffs and the Tour Championship. It rewards consistency across a whole season, but it also creates a different competitive incentive — players must weigh playing every event against saving energy for the big ones. The FedExCup points table records outcomes, not trade-offs. And the trade-off is what separates a wise player from an exhausted one in September.

Women's and amateur golf: the largest gap

If there is one area where the golf data table is systematically empty, it is women's and elite amateur golf. ShotLink data for women's events is far thinner than for men's. Major amateur events, where young players shape their careers, often carry only basic statistics. This creates a paradox: young talent is judged on the thinnest data available.

This is where I connect to a view I have long held on youth development. Young players who mature early are often pushed into adult competitive rhythms before their bodies and minds are ready, and because data on them is so thin, no one detects the signs of overload until injury strikes. The data table does not record fatigue. It records only outcomes, and the short-term outcome of an exhausted youngster can look identical to that of a rising one. That similarity is one of the most dangerous traps in the profession.

I remember a season when the pandemic emptied stadiums and cut the schedule. I had to rebuild a form-prediction model under conditions with no match data. I proposed using GPS training data from the youth team and precedents from historically disrupted seasons. The coaching staff objected. I persisted, proving the case with figures from a season after the earthquake disaster. The result: the club survived relegation, losing only two of ten restart rounds.

The lesson I kept is not "training data replaces match data". It is: when primary data is empty, seek substitute data and state its confidence clearly. Golf lacks that habit. Without ShotLink, people do not say "there is no data". They use basic statistics and present them as if equivalent.

The Vietnam–Japan lens

I was born in Vietnam and work in Japan, so I see golf's data gaps through two lenses. In Japan, a culture of disciplined practice and event infrastructure allows data to be collected fairly consistently. Hideki Matsuyama's 2026 Masters win — the first major by a Japanese male golfer — was not merely an emotional moment; it was the output of a training and record-keeping system run for decades.

In Vietnam, golf is booming in courses and players, but the data infrastructure is still young. This does not mean Vietnamese golf is weak. It means any analysis of Vietnamese golf must begin with an admission: we are talking about very little data.

I do not want to turn this comparison into decoration. I keep it only when the gap is large enough to mean something. And here it is large enough: when a Vietnamese player and a Japanese player are the same age with the same scoring average, I cannot say who is improving faster, because I do not have the same kind of data for both. The silence of the data table, in this case, says nothing about talent. It says something about infrastructure. And infrastructure can be built; talent cannot.

The counter-intuitive point: absence of data read as absence of risk

This is the part I want to give the most words to, because it is the core of everything above.

In data analysis there is a subtler trap than using wrong numbers: reading emptiness as safety. A risk table full of "no data yet" cells looks very much like a clean risk table. A model that finds no signal looks very much like a model confirming no signal exists. These are two entirely different things, and confusing them is the costliest error an analyst can make.

I once built an analysis pipeline where, when the input data was empty, the system still produced a formally complete report, with every cell filled as "insufficient information". Technically, that report was correct. In practice, it was dangerous, because a hurried reader could mistake it for a completed report and conclude there was no problem. The biggest risk was not in the data. It was in how people read the data.

What did NOT happen often tells the truth more than what did. But only when we can distinguish "what did not happen" from "what was not recorded". The first is an event. The second is a fault. Golf, with its layered and non-uniform data systems, is full of the second disguised as the first.

This is also where I must criticize myself. For years I presented my models with a confidence they did not deserve, because I did not clearly mark what was data and what was inference. I have learned that an honest analyst states not only what he knows but also what he does not know, and says it loudly. Three sentences, no more: this is the data; this is the limit; this is what I will check next. Self-criticism must have a stopping point, or it becomes a ritual of absolution rather than a method.

Correlation is not causation, and in golf this is especially hard to see

A player increases driving distance and improves his ranking. A hasty conclusion: driving distance creates ranking. But both may be caused by a third thing — a healed injury, say, or a schedule change allowing more practice. Golf is a sport of hidden variables, because each player is his own ecosystem: fitness, psychology, equipment, schedule, home course.

Strokes Gained gives us excellent correlation. It does not give us causation. When I read an analysis saying player X won because SG: Approach rose, I always ask: rose against himself, or against the tour? Over what window? On which courses? Those three questions are usually enough to collapse a pretty conclusion.

I do not believe in luck; I believe in cultivated probability. A player putting well across three straight rounds is neither luck nor talent. It is a probabilistic outcome with too small a sample to classify. The analyst's job is to say so, rather than slap on a label that sounds certain.

Signals for the next round

I will not end this with a conclusion. I will end with what I will track.

First, I will watch whether Data Golf and independent platforms publish their data-coverage levels more transparently. A model that does not state where it lacks data is a model I cannot fully trust.

Second, I will track the Ball Rollback roadmap from now to 2028, especially at the recreational level, where data is thin. If governing bodies publish real impact studies on amateur players, that will be a valuable signal.

Third, I will watch how OWGR and the majors handle the LIV-player question, because this is where a ranking system touches directly on people's livelihoods.

And fourth, I will keep logging the gaps in my own models. Not to flagellate myself, but so readers know exactly where they stand on the data map — and where the white space no one has drawn yet begins.

The golf data table will keep growing. More columns, more metrics, more models. But I believe the greatest value of an analyst in the coming decade lies not in adding columns, but in pointing out which column is still missing. The hole in the data table is where the real answer hides. And the first person to see it will not be the one with the most data, but the one most honest about what he does not have.