When the Golf Data Table Is Empty: An Analyst's Discipline of Not Inventing Numbers

A Morning in Nagoya, an Empty Table

6:40 a.m., July 14, 2026, in a small apartment in Nakamura ward, Nagoya. I reopened a data file exported from the manual shot-tracking system of an Asian Development Tour event held in central Vietnam. The file had 1,214 rows. Each row was one shot, recorded by hand by two volunteers standing on either side of the fairway and later typed into a spreadsheet. The column for distance to the pin after the shot was blank on 38 percent of rows. The column for green grass type was entirely blank.

My phone rang. A coach I work with asked: "Can you check where my player lost strokes yesterday?" I looked at the table and answered: "I don't know yet." Four seconds of silence on the other end. "You have a mountain of data right there." "Yes. And that is exactly why I don't know."

That is the hardest answer in this profession. Not because it demands technical skill, but because it runs against instinct. When a spreadsheet is in front of you, human instinct is to fill it. The brain dislikes empty cells. In golf analysis, however, an empty cell is often the only fully trustworthy piece of information.

Golf Data Infrastructure: The Vietnam-Japan Gap

To understand why I could not answer that question, we need to talk about what sits behind every number: the measurement system.

The PGA Tour put ShotLink into operation in 2026. It records every shot on site, tied to course coordinates, and from it comes Strokes Gained, which measures the advantage of a shot against the tour average in the same situation. From the mid-2010s, Strokes Gained was split into four categories: off the tee, approach, around the green, and putting. In theory, those four columns let you break a round into four separate stories.

In Japan, where I live and work, the JGTO runs its own statistics system for the Japan Golf Tour. The data is thinner than the PGA Tour's, but rich enough to track weekly trends, hole by hole and distance by distance. When Hideki Matsuyama won the Masters on April 11, 2026, the first major title by a Japanese male golfer, what stood behind that moment was nearly two decades of accumulated data about a golf nation that measures itself.

In Vietnam, by contrast, most domestic events and even Asian Development Tour events staged there, such as the BRG Open Golf Championship Danang, still record data by hand. No club coordinates, no tour baseline, no shared definition of what counts as a successful approach. Based on my experience attending tournaments and rounds on site, the biggest gap between Vietnamese and Japanese golf is not in the swing. It is in the quality of the ruler.

According to World Amateur Golf Ranking data I follow, Nguyen Anh Minh is the first Vietnamese golfer to reach the world amateur top 100. That is a milestone worth noting. But when someone asks me where his strength lies, I have to say plainly: we know the results, we do not know the causes. Between those two things lies a data gap that could be closed in three years, in ten years, or never, if the process does not change.

Four Questions Before Filling Any Empty Cell

My job is not to produce numbers. My job is to decide which cells are allowed to stay empty.

Before filling any missing value, I ask myself four questions.

One: what decision will this value serve? If it exists only to make the table look fuller, I skip it. If it affects club selection on a specific hole in a specific round, it matters and must be handled differently.

Two: am I replacing ignorance with knowledge, or with a prejudice shaped like knowledge? This is the thinnest line in the entire profession. When I insert a tour average into an empty cell, I am not filling the data. I am erasing the possibility of detecting an anomaly.

Three: if this assumption is wrong, which way does it err? Does it make the golfer look better than reality, or worse? An analyst who cannot answer that question is not an analyst yet; he is a presenter of numbers.

Four: what is NOT happening in this dataset? If a golfer hit 14 greens in regulation in round three but only one of those shots has a recorded distance to the pin, then the number 14 says very little. What did NOT happen often speaks more truthfully than what did.

These four questions cost me about fifteen minutes on a 1,200-row dataset. In a week with four rounds, that is an hour of work producing no deliverable. Nobody pays for an hour like that. That is precisely why many people skip it.

A Verified Example: Shot Number 214

I took row 214 from that morning's file to illustrate. It is a row with everything recorded: hole 9, 163 metres, 7-iron, ball on the green, 4.2 metres to the pin, one putt for the score. Looking at the row, a conclusion presents itself: this was a good approach.

But I need three more facts before I dare call it good. First, how long is hole 9 at that course and how many tiers does the green have. Second, which way the wind blew at 2 p.m. that day, because the file timestamps this shot in the early afternoon. Third, the field average for an equivalent shot, expressed in metres to the pin.

None of those three facts are in the file. I can pull wind from the nearest weather station. I can measure the hole from satellite imagery. But the field average cannot be inferred. And without an average, I have no way to know whether 4.2 metres is excellent or merely normal on that hole.

The result: row 214 was flagged "insufficient data." In the final aggregate, it sits inside the 38 percent that was discarded. That golfer lost one recorded stroke. But had I filled in an average, I would have lost the chance to discover that he is genuinely strong from 150 to 170 metres, a finding far more valuable than a filled cell.

Gaps in a data table can speak, if we are willing to listen.

Gegenpressing on the 17th Hole

There is another way to see the same problem, one I carried over from football.

In football, gegenpressing is the art of winning the ball back within seconds of losing it, before the opponent organises. Its essence is not running more, but running inside the right time window.

Golf has a similar window. After a bogey, the stretch between the tee shot on the next hole and the first putt decides which direction the round takes. If the golfer recovers rhythm across the next two holes, the bogey is an incident. If not, it becomes a streak.

Based on my experience watching tournaments on site, the metric worth tracking here is not the number of bogeys, but the average distance to the pin across the two holes immediately following a bogey. If that distance grows, the golfer is playing safer, a psychological reaction rather than a tactical choice. If it holds or shrinks, the golfer is attacking again.

I tried building this index on data from two domestic events. The problem: only 61 percent of shots after a bogey had a recorded distance to the pin. The rest were blank. The index is not reliable enough to publish.

Gegenpressing does not break the data; it breaks my assumptions. My assumption was that golf can be analysed with the same pressing logic as football. The logic holds, but the measurement infrastructure does not yet permit it. This is an example of a good idea blocked by one empty column in a spreadsheet.

The Counterintuitive Part: Analysts Are Paid to Be Confident, Not Correct

This is the part I consider most important, and the part that has cost me the most clients.

In professional sport, analysts are hired to answer. When a coach asks where his player is weak, the answer "I don't know" delivers no perceived value. A confident answer, even a wrong one, delivers value immediately. That incentive structure pushes analysts toward fabrication systematically, usually unconsciously.

The mechanism works like this. There is an empty cell. We insert an average. The average looks plausible. The table fills up. Nobody checks. Three months later, a training-plan decision is made on that average. The golfer trains the wrong thing for six weeks. No one can trace the cause, because on paper everything was valid.

Every number is a confession not yet written into prose. When I write "4.2 metres," I am confessing that I accepted a specific definition of success, a specific method of measurement, and ignored everything else. If I do not write that confession down, readers assume the number is objective. It is not. It is a choice.

The biggest blind spot in golf analytics today is not the model. It is that we blame golfers for the failures of our instruments. A golfer is judged a poor putter because his Strokes Gained Putting is negative. But if ball positions on the green were recorded by untrained volunteers, positional error can reach several metres. At that scale on a green, every conclusion about putting collapses. The correlation between a low index and poor technique is spurious, produced by measurement error.

Here is another case of correlation without causation. Golfer A posts a very high positive Strokes Gained Approach at a short course with wide fairways and soft greens. Golfer B posts a lower figure at a long course with strong wind and firm greens. On the table, A is better than B. In reality, the two are playing different sports. The golf course is an unstandardised variable, and any comparison across two different courses is a comparison between two different rulers.

I have made this mistake. In 2026, while working in football data analysis, I omitted the home-advantage factor from my model and got 6 of the final 10 matchweeks wrong. I sat through the full match footage and realised raw data was not enough; it needed tactical context. That lesson followed me into golf intact. Data is never wrong; I simply asked the wrong question.

Signals for the Next Round

If you follow Vietnamese golf next season, here are three things I will be watching.

First, the number of shots captured by devices rather than by hand at domestic events. This is an infrastructure signal, and it determines everything else.

Second, the emergence of a shared definition of distance to the pin in published data. Without that definition, no comparison across events is possible.

Third, and hardest to measure: whether anyone dares publish a table with empty cells. An organisation willing to leave gaps is an organisation still holding its discipline.

When data hides its face, error becomes the guide. Here is the question I leave open for myself this season: if Vietnamese golf could build only one measurement system in the next three years, should it measure approach shots, which decide the score, or driving, which is easiest to measure? Choosing the easy one is the shortest path to a full table. Choosing the one that decides the score is the longest path to a golf nation that knows where it stands.