The Empty Strokes Gained Sheet and the Limits of Data-Driven Golf Analysis

That evening, after the first round of a VGA Tour event, a touring professional asked me exactly one question: "What was my SG: Approach today?" I opened the laptop, scrolled down the data sheet, and the answer was right there — an empty cell. That empty cell had nothing to do with how he had played. The tournament had nobody stationed behind the fairway recording ball positions after each shot. We knew he finished three under. We did not know how many of those three strokes came off the tee, how many from approach irons, how many from the putter.

A week later, I received an eight-dimension analysis of a golf article. All eight dimensions were empty: no tournament name, no player name, not a single fact. The analysis was still presented in full, properly templated, with "insufficient information" in every field. Whoever produced it chose the most honest route — no invention. Reading it was a relief, because out there are plenty of golf analysis sheets that look beautiful, full of numbers, and contain nothing but a decorated empty cell.

On the PGA Tour, that empty cell almost never exists. ShotLink — the PGA Tour's official ball-tracking and player-tracking system — has been deployed since 2026, using optical units and volunteer crews to record the position of every shot on most holes of every round. From that raw feed, Mark Broadie, a Columbia Business School professor, published the Strokes Gained method in 2026 and put it into the book "Every Shot Counts" in 2026. The core idea is simple: every shot has an expected value based on distance to the hole and lie; the difference between the expected value before and after the shot is the strokes that shot gained or lost.

Broadie's research found that among PGA Tour players, approach shots account for roughly 40 percent of scoring differences, driving about 28 percent, the short game about 17 percent, and putting about 15 percent. That is why SG: Approach is the first number requested in any serious data meeting.

In Vietnam, we have scoring, fairway percentage, greens in regulation, putt counts. We do not have distance to the hole after each shot. Without distance, there is no expected value. Without expected value, there is no Strokes Gained. Any "skill breakdown" table for Vietnamese golfers today is an inference from total score — from outcome, not from process.

This leads to a habit I see fairly often in domestic golf coverage: fairway percentage and greens-in-regulation are cited as proof of skill. They are the outcome of strategic choices on a specific course. A player who hits driver on every par four will post a lower fairway percentage than one who lays back with a two-iron, while his remaining distance to the hole is shorter. Placing those two percentages side by side without distance-to-hole data is comparing two different strategies and calling it a comparison of skill.

For amateurs, we have the WAGR ranking. For professionals, we have OWGR. When Nguyen Anh Minh plays an international amateur event, we know where he sits in the world amateur ranking. We do not know how many strokes better or worse his iron play is than it was a week ago. Those two answers serve different purposes, and a working analyst has to tell them apart.

I hold a fairly severe professional rule: a null result is still a result, and in golf analysis the empty data cell is the most valuable information on the sheet. When all eight analytical dimensions are empty, the correct action is to stop, state "insufficient information," and attach a diagnosis of the data pipeline. The wrong action is to fill the empty cell with a story that sounds plausible.

That framework, when it has data, runs on a tight causal chain. A technical assessment missing the approach and putting splits is disqualified, because those are the two decisive metrics; without them, any comment on a swing is picture description. A form assessment without a time window is not form, because golf has a 28-to-38 peak zone and a four-round stretch is not enough to separate skill from luck. A tournament assessment without an event tier cannot compute OWGR points, and without OWGR points nothing can be said about major eligibility. A course-fit judgment without a defined course archetype — windy seaside links, soft parkland, or fast-drying desert — is just a feeling.

More concretely for the Vietnamese market: most domestic professional events sit outside the OWGR points system, so any comparison between a home winner and a player holding OWGR points belongs on two different axes. OWGR points depend on field strength — how many top-50 players entered, how dense the major-champion presence — not on the under-par number the winner posted. An 18-under win on an easy course and a 6-under win on a hard one cannot share a category, however different the leaderboards look.

Then comes the bigger stage: governance. In March 2026, LIV Golf withdrew its application for OWGR points. That is a fact with a date and a direct consequence: LIV players do not accumulate points in the ranking that decides major invitations, so their route into the four majors must pass through other doors — past rankings, invitations, or continental tours. In December 2026, the USGA and the R&A announced changes to ball conformance testing, effective January 2028 for professional play and 2030 for recreational players. Both facts belong to the same analytical layer, and both start with a verifiable question: who announced it, when, and to whom does it apply.

What I most want readers to keep: the empty cell in the data pipeline has diagnostic value. An eight-dimension analysis that comes back blank is not a thin article; it is evidence that the text-retrieval stage failed — a blocked source, a JavaScript-only page, or a truncated original. If the pipeline had a hard validation gate — refuse to proceed when the fact list is empty, or when the title is blank — the entire downstream output would never have existed. A mature data practice is not measured by how many dashboards it produces, but by how many times it dares to return "no."

The place where analysts slip most easily is the step from "no data" to "there is a story." Correlation is not causation, and in golf that temptation dresses very politely. A player putts well for four straight rounds, the numbers jump, and the commentary declares he has "found his touch." Four rounds is roughly 60 to 70 putts; that sample is far too small to separate from the natural variance of putting skill itself. The same happens with approach metrics: one week the irons finish next to the flag, the next week they find the fringe, and the narrative flips while the underlying skill has not moved.

Numbers do not lie. But reputation whispers into the ear of the person who does not read the sheet. A player praised for "hitting it miles" may be losing strokes in exactly that area once his driving is placed against the tour baseline; a player criticised for "poor putting" may simply have met a tougher stretch of greens than average. Both conclusions require the same three things: shot-level data, sample size, and course context.

I also have to be blunt about the risk in my own trade. When input data is insufficient, the risk of a wrong conclusion is very high — I put it above 70 percent for any technical judgment built from total score alone. My threshold for publishing analysis is below 20 percent. Beyond that threshold, what is Plan B? For Vietnamese golf, Plan B is very concrete: record by hand. Two volunteers, one at the landing zone, one at the green, logging distance to the hole with a laser for three consecutive events. At roughly 400 shots per event, we would have enough sample to build a rough expected-value model for Vietnamese course conditions themselves, instead of borrowing the PGA Tour baseline and applying it to turf shaped by monsoon wind and completely different green speeds.

I do not predict. I read the data and accept the consequences. But I also know data does not speak for itself: it needs someone to build the pipeline, install the validation gate, and be willing to leave the empty cell alone until someone fills it with a real measurement. I started a blog from a lecture hall, believing data would speak on its own. Eleven years later, I teach it to speak in sentences.

The signal to watch in the next round: which domestic tournament becomes the first to publish distance to the hole after every shot. That is the cheapest indicator that a golf market is moving from storytelling to measurement.