On a Tuesday afternoon at a tournament in Florida, in the stretch of the property where the television cameras never go, a player stood on the practice green and hit 40 putts from roughly 12 feet. He made 31. His caddie said something I copied into my notebook word for word: "This week he's the best putter on the planet. Next week the broadcast will call asking why his number dropped more than a stroke a round."

Three weeks later, that is exactly what happened. His Strokes Gained: Putting fell from +1.4 to -0.3 per round. No swing change. No injury. No new putter. Different greens, different wind, and a 72-hole sample that the media read as though it described the man's character.

I have covered professional golf for 21 years, and most of that time I have read the data sheet before I read the swing. What bothers me most every regular season is not wrong numbers. It is correct numbers used in the wrong place.

Context: a metric built to fix an old error

In 2026, the PGA Tour began recording every shot through the ShotLink system. Ten years later, from 2026, the tour published the Strokes Gained family officially. The person who formalised it into a comparable measure was Mark Broadie of Columbia Business School, and his 2026 book "Every Shot Counts" pushed the concept into the everyday language of golf.

The core idea is simple. Every position on the course carries an expected number of strokes to finish the hole, drawn from tour-wide averages. A putt holed from 12 feet earns more credit than the expectation; a putt missed from 12 feet costs less than leaving the ball three feet from the hole. This replaced older measures such as putting average and greens in regulation, which could not tell a three-foot putt from a thirty-foot putt and therefore blended two entirely different skills into one cell.

Strokes Gained fixed that, and fixed it well. It lets you compare a player hitting an approach from 150 yards with one hitting from 180. A traditional scorecard cannot do that. But a good measuring tool does not automatically produce a good conclusion. This is where golf media is stuck, and it is where I spend most of every season re-checking the work.

Core: four days cannot measure a craft

Every week the stat board offers four headline numbers: SG: Off the Tee, SG: Approach, SG: Around the Green and SG: Putting. Add them up and you get SG: Total. Media takes that week's SG: Total, compares it with last season, and writes a story about progress or decline. The problem is that four tournament days produce only about 250 to 300 shots, of which perhaps 115 to 125 are putts — a sample far too small to separate skill from luck.

The standard deviation of SG: Putting for a PGA Tour player is roughly one stroke per round. A +1.5 week and a -0.5 week can come from the same player, with the same technique, the same putter, the same stroke. If you watch one tournament, you are not measuring skill. You are measuring noise. That is why SG: Putting correlates far more weakly year over year than SG: Approach, and why Broadie's analysis shows most of the scoring difference between elite players comes from driving and approach play, with putting accounting for only a small share.

Put differently: the data sheet tells you clearly who hits approach shots well, and only vaguely who putts well. Broadcast coverage does the opposite. Every week the lead character of the show is the man with the highest SG: Putting, which is the most reversible of the four numbers.

A single season is one sentence in a book that runs a decade long.

Then come the maps. Dispersion maps and heat maps flood the coverage, drawing a cloud of shot endpoints for each player on each hole. They look scientific, and sometimes they genuinely help: a cloud shifted hard right across four days is a signal you cannot ignore. But a heat map only tells you where the ball finished. It does not tell you where the player was aiming. Two shots 25 yards apart can look identical on the map, when one was a miss and the other was the correct execution of a plan into a pin tucked in a dead corner. The second shot looks worse and plays better. The map cannot tell the difference.

Caddies hold the information the map is missing. During an untelvised practice round, I once asked a veteran caddie how he read a hole with the pin on the upper right tier. He said: "I don't care where the ball lands. I care whether he commits to the shot he picked." That note appears in no statistical table, and it explained four days of golf better than four metrics did.

I have also built my own tracking sheet for a tournament swing through Asia: how many times a player changed club after taking his stance, how many times the caddie was called back for a second read, how many times the player read the green alone. Those three variables together predicted final-round performance better than the previous week's SG: Putting. Not because my variables are cleverer. Because they are not contaminated by whether the ball happened to find the hole.

Based on my experience tracking matches, one thing has repeated often enough for me to trust it: the players who hold high standing across multiple seasons are not the ones with the prettiest SG: Putting every week, but the ones with the narrowest gap between their best and worst weeks. Consistency is legible in three years of data. It is invisible in three weeks.

There is a deeper layer. A data board does not merely describe golf; it underwrites a market. A 20-year-old who wins once after 54 good holes is immediately labelled the next generation, invited to events with sponsor exemptions, and placed on endorsement lists before he has played a full season. This industry has learned how to turn a small sample into a sellable product, and the junior ranks are where it sells hardest. A golf academy in the United States charges somewhere between 30,000 and 60,000 dollars a year per student; behind that tuition are families relocating, borrowing, and staking a 14-year-old's entire future on 54 holes at a junior event. It is the most sophisticated lottery ticket I have seen in sport. PGA Tour membership cards are few, and Korn Ferry Tour promotion slots are fewer. The story that sells the ticket has no limit at all.

Contrarian: the industry is betting on the wrong kind of data

The thing being left out sits outside any fifth or sixth metric. It is everything nobody records: practice rounds, caddie conversations, the wind at seven in the morning, the pin position on Saturday, and whether a player believes in his own plan.

In June 2026, when the PGA Tour returned at Colonial without spectators, I had a chance to observe something rarely observable: players competing with no energy arriving from the galleries. Tempo changed. Some players became more mechanical; others lost an invisible acceleration that crowd noise supplies. When the stands are empty, the match exposes what tactics conceal. Read only the stat sheet afterwards and you will see nothing unusual. That is precisely the problem.

In a press room, I was once cut off for asking about tactics instead of about a star's family life. I did not argue. I went away, built a tracking sheet, and let the numbers speak. They doubted the voice before they heard the argument. I learned to secure the evidence first and the expectation second. The same rule applies to golf data: evidence before story, never the reverse.

This is the hardest part of the job. When a data sheet is empty, the professional move is to stop and say there is not enough basis — not to fill the gap with a plausible-sounding hypothesis. I have watched too many columns get populated with guesswork, and three months later nobody remembers who invented them. An honest blank is worth more than a filled-in error. Coldness is a long-term strategy, not a character defect.

What to watch

Golf will keep producing more metrics. There will be more cameras, more sensors inside shafts, more biomechanical data and even heart-rate data on the greens. The question is not how much we will measure, but whether we will admit the limits of what we measure.

Next time a stat board tells you a player has found his form, ask two things: how large is the sample, and what was left off the board. The answer usually sits in the stretch of the property where the cameras never go.