The Blank Strokes-Gained Table: Where Golf Analytics Keeps Fooling Itself

On a Monday morning I opened a spreadsheet and counted four columns: off the tee, approach, putting, total. The headers lined up. The number format was preset to two decimals. Pale grey gridlines ran evenly down the sheet. Under every heading sat a blank space.

That table passed every structural check the system imposes. All fields present. Correct order. Correct formatting. Structural integrity score: one hundred percent. Content: nothing.

I stared at it longer than necessary. Thirteen years working with sports data taught me one thing: the most dangerous object in this trade is a properly formatted, hollow table. A wrong table can be argued with — you can hunt down the error. A blank table with full headers looks exactly like a complete one to anyone without time to open every cell. In golf, almost nobody has that time.

Golf is the most heavily measured sport I have worked in, and the one readers can least verify for themselves.

The PGA Tour's ShotLink system records the coordinates of nearly every shot at nearly every hole across sanctioned rounds. From that dataset Mark Broadie, a Columbia Business School professor, built strokes gained, published it in "Every Shot Counts" in 2026, and the PGA Tour adopted it as an official statistic that same season. The logic is compact: each shot is compared with the tour average from the same distance and the same lie. A positive differential means better than baseline, negative means worse. The metric splits into four buckets: off the tee, approach, around the green, putting.

The split is not decoration. Broadie's work shows the long game — driving plus approach — explains most of the scoring difference among tour professionals, while putting is the most volatile bucket week to week. That is why any serious analyst weights approach before opening their mouth about putting. One hot putting week forecasts nothing. One strong approach season forecasts plenty.

But the unit of strokes gained is abstract. Nobody sees "plus 0.43 strokes". Nobody at home can distinguish an approach 0.2 strokes better than baseline from one 0.4 better, by eye, on a phone screen at two in the morning. In football, expected goals has a physical referent: you see the shot from the penalty spot, you see the keeper, you see the gap. In golf, the only referent is the table.

Which means the entire trust of a golf reader funnels into one place: whether the writer is honest. There is no second check.

Here the structure of golf data betrays itself. A real strokes-gained table and a fabricated one are indistinguishable in feel. Both have four columns, plus and minus signs, two decimals, and fit in one glance. The reader has no way to tell. Even a professional writer, without returning to the source, cannot tell.

I tested this on myself. In 2026 I built an xG model in Excel to analyse 26 rounds of V.League and wrote a piece on Quang Nam FC — champions with 48 percent average possession, lowest of the top five, but a 17.5 percent conversion rate. The piece was mocked. Three months later Quang Nam won the title. The lesson I took was not "I was right". The lesson was: when a table looks plausible, people assume it is correct — even when it is wrong, even when it is empty.

Golf carries far more risk. The schedule never stops. One event a week, one article per event, one new metric per article. The content machine does not wait for data. It waits for a string of characters that resembles data.

I have watched that process from inside a newsroom. When the source body fails to load — paywall, JavaScript page that never rendered, a video with no transcript, a URL that resolved to a section front instead of an article — what remains is an empty article skeleton with every heading intact. That skeleton looks like an analysis waiting to be filled. Someone always fills it.

This is why a blank golf data table pushed downstream is the most serious failure in the whole chain, worse than a missing article. A missing article gives you nothing to read. Filling the gap gives you a smooth, professional, entirely untrue piece.

That Monday spreadsheet gave me exactly one datum: the domain label said "golf". Everything else — headline, source, date, information points, entities, author stance — was empty. In the eight-dimension audit I use to read a golf piece, seven and a half dimensions had no input.

One detail mattered. The domain label still resolved. The pipeline could still see the page's metadata layer — headline, byline, timestamp, tag set — even though the body was unreachable. That is a small technical signal with large practical value: recovering only the headline and publication date would have fixed the tournament, the event tier and the timeliness. Those three are enough to write something useful. No strokes-gained line required.

Numbers do not lie, but they only speak about the course that produced them.

I say this having stood on the other side. In 2026, when stadiums closed, I was a data assistant at Becamex Binh Duong and found the home win rate in V.League fell from 49 percent in 2026 to 38 percent without crowds. The coaching staff wanted to keep the home-away game plan. I objected, presented a 42-match comparison, and proposed active defending away from home. The team won four of the next five.

The lesson I carried into golf was not "data wins". It was that context variables — crowd, surface, weather, fixture density — can reverse a conclusion that was correct elsewhere. I hate uncertainty. But 2026 taught me an unforecast variable can be stronger than any algorithm.

In golf that variable is usually the course. A player with plus 0.8 approach strokes gained on tour can go negative on a seaside links with hard fairways and heavy wind. Nothing is wrong with the number. It was simply measured elsewhere. Dropping it into a prediction for a links course without a note is the worst misuse of data: right number, wrong place.

On volatility, be blunt. Three good putting rounds is an anecdote, not a trend. How many rounds before putting has genuinely improved? The honest answer is more than one tournament, and even then most of the gap sits inside the noise band. Serious analysts rank putting last among the buckets permitted to explain a result — after approach, after driving.

News desks do the reverse. Putting is the easiest thing to narrate. "He putted like a man possessed over the last nine" is a story. "He gained 1.2 approach strokes in crosswind at the 14th" is not a story. It is data. Data does not narrate itself.

That is my job. I started a blog from a lecture hall, believing data would speak for itself. Eleven years later, I teach it to speak in sentences.

There is another layer Vietnamese golf readers rarely see: the Official World Golf Ranking is a governance arrangement, not a physical constant. In October 2026 the OWGR declined to award points to LIV Golf events, citing a format that did not meet criteria. That decision said nothing about the quality of the players. It said something about who gets to define a qualifying event.

When you write "this player is 47th in the world", you are citing an agreement, not a natural fact. In a news brief the distinction is trivial. In analysis, it is the whole distinction.

Equipment rules behave the same way. The USGA and the R&A announced restrictions on ball flight distance on 6 December 2026, applying to professional and elite play from January 2028 and to recreational golfers from January 2030. A piece that does not separate the professional impact from the weekend-golfer impact is wrong in its premise — even if every figure quoted is accurate.

There is a further blind spot golf models have never solved: the bag and the person carrying it. A player changes swing coach, changes caddie, changes equipment contract — no strokes-gained field records any of it. When the numbers dip six months later, the article calls it a form slump. It may have been an equipment switch that never settled, or a caddie relationship that frayed.

This is where the industry's valuation models — in football and golf alike — fail in the same direction: they overrate young potential and underrate environmental chemistry. A 21-year-old talent is priced at the peak of a curve nobody has tested. A caddie who has read four hundred rounds alongside his player is priced at nothing. But when a 2.5-metre putt on the 18th decides the week, the person reading the green is the variable.

Age has a curve too, even if few bother to draw it. The competitive peak for a tour professional usually sits between 28 and 38, with a tail past 40 for those who hold fitness and putting technique. A 22-year-old being celebrated to the skies has not yet entered his best years. Meanwhile junior schedules thicken, and bodies that are not finished developing get pushed into adult rhythms. Back and wrist injuries under 25 are not rare.

The media story around two leading players is the cleanest illustration of the gap between data and legend. Rory McIlroy has spent multiple seasons among the PGA Tour leaders in driving distance — a measurable skill, no argument needed. Scottie Scheffler rose as the best tee-to-green player in the world while putting was the weakest part of his profile, and he won anyway. Two entirely different profiles, and both get told wrong when folded into a single story about "form".

A metric only means something with the course, the weather and the point in the season attached. Strip those three away and you have a handsome string of characters.

In Vietnam, data-driven golf writing carries an extra layer of difficulty. Primary sources barely exist domestically. Most information arrives via translated reports, social media or tournament press releases. Every intermediary layer strips context from a metric. After three layers it becomes a floating figure with no root.

I have spent enough mornings at a practice range in Binh Duong to know that a player's feel and a player's data often disagree. Some strike the ball beautifully, the flight is pure, and they still drop shots because they chose the wrong target. Some strike it badly and still make par because they know exactly where they stand on the course. No table records the second part. If I write "he is striking it well" based on the sound of contact, I have fabricated a metric out of feeling.

Vietnam's first recognised golf course opened in Binh Duong in the mid-1990s. Thirty years later the country has hundreds of courses, a domestic professional tour, and a golf-tourism flow into Da Nang and Hai Phong. The data infrastructure lags the course infrastructure badly. No ShotLink, no strokes gained for domestic events, no open data repository for cross-checking. Every piece of domestic golf analysis therefore rests on direct observation or imported data — and both can be wrong.

This is why I refuse an unsourced data table. Not out of strictness. Because I have no tool to verify it.

But here I want to argue against myself, and against most of the industry's reflexes.

An empty technical dimension is not necessarily a failure. Some golf pieces are professionally valid without a single strokes-gained figure: pieces on tournament structure, sponsorship contracts, course development, disputes between tours. For those, an empty technical dimension is the correct answer, not missing data. Forcing a strokes-gained table into a piece about tour restructuring manufactures false precision.

The real problem is not that machines invent numbers. The real problem is that the industry's verification habits were never built to catch invented numbers. An article with twelve metrics gets edited more carefully than one with two, because it looks "evidence-based". Metric count becomes a certificate of quality. The more numbers, the fewer questions.

Meanwhile correlation keeps getting read as causation. Driving distance up, scores down — sounds right, except short courses and wide fairways move both at once. Good putting and victories — sounds right, except good approach play creates short putts, and short putts flatter the putting number. The causal chain runs opposite to the story being told.

I do not predict. I read the data and accept the consequences.

And sometimes the data reads: there is nothing to read. That sentence sells no advertising, trends nowhere, gets shared by no one. But if a newsroom learns to publish it — flatly, without apology, with the technical reason attached — every metric that appears afterwards gains a level of credibility. In a golf market growing faster than its data infrastructure, that is the cheapest remaining investment.

Next time you read a golf piece with a handsome strokes-gained table, ask one question: where did that table come from, and who checked it. Numbers do not lie. But reputation whispers into the ear of the person who does not read the table.