Eight Headings, Zero Data Points: The Structural Vacuum in Professional Golf Analysis

Nha Trang, March. The air conditioning in the meeting room is running too cold. On the table sits a twenty-two-page document: a logoed cover, a Roman-numeral table of contents, eight chapters. Chapter one, technical analysis. Chapter two, player and form. Chapter three, tournament system. Chapter four, landscape and governance. Chapter five, rules and equipment. Chapter six, risk surface. Chapter seven, media narrative and expectations. Chapter eight, industry transmission.

Every chapter has a table. Every table has a bold header row. The left column reads "metric", the middle reads "assessment", the right reads "comparison target". Not a single cell contains a number.

The presenter opens: "We have conducted a comprehensive review." The room nods. I turn to chapter one and look for the SG: Approach line — the metric group most strongly correlated with scoring among the four Strokes Gained categories. That cell reads: "to be updated."

I close the document. In eleven years of working with golf data, I have never met anything more dangerous than a report that looks finished.

Context: eight links in a chain, and the first one

The PGA Tour has operated ShotLink since the early 2000s and deployed it fully from 2026. The system records the coordinates of every shot by every player, on every hole, in every round. Mark Broadie, a professor at Columbia Business School, used exactly that kind of data to build Strokes Gained, publishing the method in Every Shot Counts in 2026. The PGA Tour made SG an official statistic in 2026. Around 2026 Data Golf appeared as an independent platform, letting people outside the tour revisit much of that data architecture.

The production chain of a professional golf analysis has eight links, joined in strict order: shot → coordinates → segment classification (off the tee, approach, around the green, putting) → normalization against the tour baseline → Strokes Gained → forecasting model → recommendation → decision.

The notable part is this: if the first link is empty, the seven that follow can still be written out. A table does not know that it is empty. A blank cell and a cell reading "not yet updated" look identical on paper, and equally serious in a meeting where someone is pouring water.

The eight chapters in that document map onto the industry's eight standard analytical dimensions: technical and data; player and form; tournament system; landscape and governance; rules and equipment; risk surface; narrative and expectations; industry transmission. The framework is correct. The framework is complete. The problem is not the framework.

I have tracked domestic golf for three years, and spent two years before that doing data for regional events in Southeast Asia. At most regional tournaments there is no ShotLink equivalent. No shot coordinates. No SG. Caddies write scores in pencil on paper cards, and paper cards never enter any database. When raw data is scarce, the natural reflex of a working analyst is to build the frame first and fill in the numbers later. That reflex is right on deadlines and wrong on epistemology: it produces documents polished enough to be trusted and empty enough to be unverifiable.

Dissecting eight empty chapters

Read those eight chapters as eight questions, then ask what data each question requires before it can be answered.

Chapter one, technical analysis. The question: which segments is this player strong and weak in. The data required: the four SG categories (Off the Tee, Approach, Around the Green, Putting), greens-in-regulation rate, driving accuracy, scrambling rate, proximity to the hole, the course's average green speed. Without the four SG categories there is no technical analysis, only description.

Chapter two, player and form. The question: where is this player on the career arc. The data required: OWGR rank and trend, tour tier, recent results with sample size, major top-10 count, cut-made rate, injury status, schedule load.

Chapter three, tournament system. The question: how much is this event worth. The data required: event tier, field strength, OWGR points scale, purse, cut line, calendar position. For team events such as the Ryder Cup or Presidents Cup, add the selection mechanism: how many automatic picks, how many captain's picks, and which criteria govern the captain's picks.

Chapter four, landscape and governance. The question: who controls the distribution of resources. The data required: the PGA Tour–LIV Golf relationship, the role of Saudi Arabia's Public Investment Fund, the standing of the DP World Tour, OWGR's recognition status. This is the chapter every empty document likes to write at greatest length, because it sounds strategic.

Chapter five, rules and equipment. The question: which regulatory framework is changing, and how many yards does that translate into. The data required: Model Local Rule G-10, published by the USGA and the R&A on 14 March 2026, effective for elite competition from January 2028 and for recreational golf from January 2030.

Chapter six, risk surface. The question: what can break, with what probability. The data required: at least one observed event, and one time frame.

Chapter seven, narrative and expectations. The question: what is the crowd mispricing. The data required: odds, expert predictions, the prevailing media frame, and the phase of the heat cycle.

Chapter eight, industry transmission. The question: where does an upstream change flow downstream. The data required: course economics, equipment brands, broadcast rights, data and betting, the talent pipeline.

Eight questions, eight different datasets. In that document, all eight were answered in prose, not in numbers. And prose cannot be verified.

How Strokes Gained actually works

To see why a blank cell in chapter one is serious, you have to understand what SG measures.

Strokes Gained does not measure a beautiful shot. It measures the expected value a shot takes away or leaves behind. For every ball position on the course — how far from the pin, in the fairway or the rough, on the green or in a bunker — the system knows the average number of strokes a tour player needs to finish the hole from there. The difference between the expectation before the shot and the expectation after it, minus the one stroke used, is that shot's value. Summed over a round and split by segment, it yields the four SG categories.

Two consequences follow directly. SG is a relative metric, dependent on each season's tour baseline: a player with unchanged technique can see SG rise or fall as the baseline shifts. Which means any report citing SG without stating the season and the comparison baseline cannot be verified.

And it also means a blank "SG: Approach" cell is not a minor formatting omission. It is the absence of all of chapter one.

What a compliant report looks like

A compliant technical report needs only three lines, but those three lines must carry numbers.

Take the 2026 season as a reference. Scottie Scheffler finished the year ranked world number one, won seven PGA Tour events including The Masters on 14 April 2026 at Augusta National, and took Olympic gold in Paris. The backbone of that season was SG: Approach. For months, his putting was framed in the media as a weakness; the data showed the largest gap between him and the rest of the tour lay in the approach segment. Xander Schauffele won the PGA Championship on 19 May 2026 and The Open on 21 July 2026 with a nearly opposite profile: consistency across all four segments, dominance in none.

The 2026 season offers a cleaner comparison. Rory McIlroy completed the career Grand Slam at The Masters on 13 April 2026, after years of being read as someone who could not close a major under pressure. Scottie Scheffler won the PGA Championship on 18 May 2026 and The Open on 20 July 2026. J.J. Spaun won the U.S. Open on 15 June 2026 at Oakmont, a case where approach and scrambling data explain most of the outcome and the courage narrative explains none of it.

These three examples share one property: they can all be checked. Anyone with access to ShotLink or Data Golf can open the table, pick the window, and confirm or refute. That is the entire standard for a data report. Not length. Not chapter count. Refutability.

I once built a report the other way round. In 2026, working as a data assistant for a football blog in Nha Trang during the World Cup in Russia, I manually logged 1,240 dangerous situations and computed xG for each one. In the France–Belgium semi-final I produced xG of 1.2 for France and 1.8 for Belgium, meaning the scoreline did not reflect the match. The editor dismissed it with a remark about gender. I wrote a two-thousand-word rebuttal with charts and posted it to a forum; it was shared more than three thousand times. That piece won, but it won because it had numbers. Had I written only "Belgium played better", it would have been dismissed a second time.

People watch the putt drop. I watch the ball flight before the putt.

Notes from a manual data build

Based on my experience tracking matches at regional tournaments, manual data collection has its own logic, and that logic is worth setting down because it is the antidote to the structural shell.

A typical regional event has around 120 players, four rounds, 72 holes. If you record only four metrics per hole — drive position, distance to the pin after the second shot, green result, putt count — one person can cover roughly three groups per round. Three groups times four rounds times eighteen holes yields about 216 data rows per week. A ten-event season yields more than two thousand rows.

Two thousand rows is not ShotLink. But two thousand rows is enough to compute a crude version of SG: Approach: if player A hits the green from 150 yards while player B from the same distance finds a bunker, the expected-strokes gap between those two positions is a real number. Multiply it across repetitions and it becomes a trait. And a measurable trait beats eight bold headings.

What I learned from that work was not the calculation technique. It was the discipline of accepting a small sample. Two thousand rows are enough to describe a trend, not enough to describe an individual player. That boundary must be written into the report, immediately beneath the number. Removing that caveat is the first step in turning an honest analysis into a well-populated structural shell.

The one-number rule

From that point I applied a single rule to every document leaving my desk: no heading may stand alone without at least one number, one source, and one date.

The rule sounds simple, and it destroys about seventy percent of first drafts, including mine. Auditing a report takes ninety seconds.

Read the table of contents. Count the chapters. Count the verifiable facts. If the chapter count exceeds the fact count, stop — the document describes a process, not a result.

Open three lines at random. If none carries a number, close the file.

For three lines that do carry numbers, ask three questions: where does the data come from, what date was it recorded, and what is the sample size. The third matters most and is skipped most often. A player leading the tour in SG: Putting over four rounds says nothing about the next four months.

The contrarian angle: the framework is not wrong, the instinct to present it is

It would be easy, and wrong, to conclude that the eight analytical dimensions are redundant. They are not. It is the best framework this industry has, and I still use it for every player file. The framework does not generate the error. The instinct to present the framework before filling it does.

The mechanism deserves naming. A document with eight bold headings creates the impression that an evaluation has taken place. Readers skim the structure and infer the content. In a meeting, that impression outweighs the actual emptiness. A blank page is safe, because everyone can see it is blank. A page with eight headings and no numbers is dangerous, because it looks finished.

This is the risk category I rank most severe: process risk, not domain risk. People do not lose because the model is wrong. They lose because a decision was made on a document that was never filled in.

There is another layer, about how golf data gets misread at the media level. The story that sells is the putting story. A hot putter is a headline. But SG: Putting is the most volatile category year over year; its season-to-season correlation is markedly lower than SG: Approach. When a player wins a major on the greens, the media calls it the cause; the model calls it variance. Both are true in the past tense, but only one forecasts the future.

In the opposite direction, the talent market makes a symmetric error. Selection models overvalue young players — because youth is a continuous, quantifiable variable that is easy to sell to a board — and undervalue what cannot be captured in a numeric column, such as the relationship between a player and his caddie. In golf, the caddie is the most underpriced variable in the entire file. I have never seen a transfer model with a column for it.

The crowd applauds on emotion, but the data hears a different rhythm.

A rare natural experiment

In June 2026 the PGA Tour returned to competition at Colonial Country Club without spectators. For several weeks the tour operated under conditions that can barely be reproduced: the same players, the same course, the same rules, but with an entire interactive dimension removed.

For a data analyst, that is a natural experiment. Shifts in pace of play, in club-selection behaviour, in how players handle pressure on the closing holes all become observable once a variable is deleted from the equation. An empty golf course does not lack applause; it lacks a dimension of data — and the missing dimension is precisely what needs measuring.

The lesson for structural shells is direct: when a data dimension disappears, a table can keep the same number of rows. Readers will not notice the missing dimension unless the writer states which one is absent. That is why every blank cell in my reports must state the reason it is blank, rather than being left white.

Industry context: three signals to track

The current major season compresses emotion to a high degree, and that makes structural shells more attractive than usual. As time pressure rises, people tend to present the frame instead of waiting for the numbers. Three signals are on my board, and all three are verifiable.

First, Model Local Rule G-10 from the USGA and the R&A, published on 14 March 2026, effective from January 2028 for elite competition. It is the first time in decades that the equipment framework has been tightened globally. The direct modelling consequence: the relative value of driving distance declines, and the relative value of accuracy and approach increases. Any player-valuation model still using pre-2028 weights after 2028 will misprice.

Second, the governance structure between the PGA Tour and LIV Golf. The framework agreement announced on 6 June 2026 has not yet closed into a clear operating structure. OWGR declined to award ranking points to LIV Golf in October 2026. As long as the pathway into majors through OWGR remains shut to a group of players, every model forecasting major fields is missing a variable.

Third, data quality at regional level. This is the least discussed signal and the most important for Vietnamese golf. Without ShotLink you can still do data — manually, by standardizing scorecards, by building a small but consistent database. What you cannot do is skip that step and then write eight chapters.

Risk surface, ranked

If I had to draw a risk matrix for the analytical process itself, it would have six rows.

Competitive risk: misjudging a player's strongest segment. Medium probability, high impact, mitigated by mandating the four SG categories.

Psychological risk: reading one hot putting week as a durable skill. High probability, medium impact, mitigated by sample-size testing.

Injury risk: omitting physical status from the file. Medium probability, high impact.

Career risk: overpricing young players against their actual development curve. High probability, high impact.

Governance risk: forecasting major fields without accounting for tour pathways. Medium probability, medium impact.

And the final row, the most severe: analytical risk — making a decision on an empty evidence base. High probability, high impact, and the only mitigation is refusing the document before it is used.

What I take from this

I write the report, close the file, and the market reopens on its own. A report sitting in a drawer is a chart waiting for a time axis.

The twenty-two-page document in Nha Trang in March will not become data by itself. It becomes data only when someone returns, opens ShotLink or opens the handwritten log, and fills in each cell. That takes three weeks. It always takes three weeks.

The worry is not that an empty document gets caught. The worry is that it does not, and a decision is made on the strength of eight bold headings.

The next data cycle opens when Model Local Rule G-10 takes effect, when the governance structure between the tours resolves, and when a new generation of caddies starts recording what they see. Whoever has the columns ready before the market reopens will be the one pricing it correctly.

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