Strokes Gained: Approach — the Decisive Variable Every Golf Leaderboard Hides

In the last three rounds of a Japan Golf Tour event, the champion averaged 1.74 putts per round, ranking 42nd in the tournament's putting statistics. In Strokes Gained: Approach, he led the field at +2.31 strokes per round, nearly 0.8 ahead of second place. The leaderboard records only the final number after 72 holes. It does not record who produced that number, how, or under what pressure.

I reopened the entire final-round footage to verify. The decisive putt on the 18th hole measured 4.2 metres, exactly the distance where the tour's success rate sits near 38 percent. He made it. But stopping there would mean ignoring the previous 17 holes, where he repeatedly placed the ball within 3 metres of the pin using irons with an average proximity of just 8.4 metres. That is the real structure of the victory.

I have covered professional golf in Japan for eight years, since I worked as a data contributor for a sports outlet in Nagoya. Before that, my background was in football data analysis: xG, PPDA, pressing sequences. When I moved to golf, I carried an old habit with me: never trust a number that stands alone.

The trouble with golf is that the leaderboard carries too much psychological weight. A made putt on the final hole creates a moment, and that moment dominates the viewer's memory. But golf is a sport of 72 holes, not of one hole. It is exactly like football: a goal in the 90th minute cannot tell the story of the previous 89.

In golf, the Strokes Gained system splits the game into four skills: Off the Tee, Approach, Around the Green, and Putting. Each metric compares a player with the tour average under the same conditions. This is a tool I trust, but it is also the tool I doubt most, because it is so easily misread.

What I want to do in this piece is not to praise SG. I want to ask the reverse question: if SG: Approach matters so much, why do some champions still win with putting, and why do players with elite SG: Approach still fail to win?

Start with the data. In the most recent Japan Golf Tour season for which I have complete data, I took 20 players with a minimum of 40 rounds to ensure the sample was not too small. I ran the correlation between end-of-season ranking and each SG metric.

The result: SG: Approach had the strongest correlation with end-of-season ranking, at r = 0.71. Next came SG: Off the Tee at r = 0.58. SG: Putting reached only r = 0.34, far lower than viewers usually assume. SG: Around the Green sat at r = 0.29.

This is the point I want to stress: putting, the skill viewers remember most, is the variable with the weakest correlation to long-term performance. But this is where I have to be careful. A weak correlation does not mean putting is useless. It means putting fluctuates more from week to week, so it explains the result of a single good week better than the result of a whole season.

I verified this backwards by splitting the data into two halves of the season. If SG: Approach is truly stable, it should predict the second half from the first. The result: the correlation between first-half and second-half SG: Approach was r = 0.64. For SG: Putting, that figure was only r = 0.21. This matches a principle I learned from my own mistake: the more a skill depends on small movements and timing, the harder it is to predict.

But here a paradox appears. When I filtered the 10 rounds in which a player had SG: Putting above +2.0, their win rate in those very rounds was very high. In other words, putting is the decisive variable in the short term, while approach is the decisive variable in the long term. Two different questions, two different answers.

"Data is never wrong; I just asked the wrong question." The question "which skill matters most in golf?" is a bad question, because it does not specify a time frame. When I changed it to "which skill best predicts end-of-season ranking?", the answer became clear and verifiable.

I should explain how SG: Approach is calculated to avoid misunderstanding. The system takes the ball's position before the shot and its position after, then compares both with the tour's average expectation from those positions. The difference is the number of strokes a player gained or lost against the baseline. The problem is that this baseline is built from a large database, often hundreds of thousands of shots. If that database does not represent the course conditions of a specific event, the metric drifts.

For example, a course in the mountains of Japan with fast greens and strong wind will have a very different expectation level from a damp coastal course. When I applied the standard PGA Tour model to Japan Golf Tour data, I found a systematic error of about 0.15 strokes per round in the approach metric. This is not a large number, but it is enough to shift the ranking of players who sit close together.

On the sample: 40 rounds sounds like a lot, but when I isolate the approach metric week by week, each week has only four rounds. The standard error of an SG metric over four rounds is about 0.9 strokes. This means a player can fluctuate by nearly one stroke per round purely by chance, with no real change in skill. This is why I never conclude anything about a player's form after a single event.

The next thing I wanted to test was context. A high SG: Approach can come from two very different sources: shorter proximity after the iron shot, or better putting from the same distance. These two sources demand two completely different training methods.

I separated proximity data from putting data inside the 3-metre range. In the group of 20 players I tracked, those with top SG: Approach averaged 8.6 metres of proximity from the fairway, against 10.9 metres for the rest. A 2.3-metre gap sounds small, but in golf, each metre closer to the hole is worth roughly 0.08 expected strokes. Multiplied by 18 holes per round and four rounds per week, that is more than five strokes per week, enough to change an entire ranking.

This is where I borrow a concept from football. In football, gegenpressing is the art of winning the ball back immediately after losing it, at the closest distance to the opponent's goal. In golf, there is a similar mechanism: after a tee shot into the rough, a good player does not try to "save" it with a miracle shot. They accept a safe approach to the green, then use the first putt to create pressure. I call it gegenpressing in golf: recovering position from a bad spot by limiting the damage, instead of trying to win everything back in one shot.

But I must confess: the first time I used this comparison, I used it wrongly. In 2026, I wrote a piece comparing golf's rhythm with gegenpressing without a single number to back it. I just felt the two things "sounded alike" and wrote. That was the mistake of a storyteller, not an analyst. "Gegenpressing does not break the data; it breaks my assumptions." Since then, I have set a rule for myself: every cross-disciplinary comparison must come with at least one metric proving the similarity, or it is cut.

And what is the number that proves it here? When I measured the par-recovery rate after a bogey, that is the rate at which the next hole after a bogey is played at par or better, the top SG: Approach group reached 68 percent, against 54 percent for the rest. A gap of 14 percentage points. This is evidence that approach skill is not only about hitting the ball well, but also about controlling emotion after a mistake, because a good iron shot is the tool for regaining rhythm.

Japanese players such as Hideki Matsuyama and Takumi Kanaya have long been known for elite approach play. But what I want to stress is that even for players with the best approach, results can fail to match in any given week, because putting can collapse or explode. This is the proof of the short-term and long-term paradox I described above.

My experience following matches gives one observation that aggregate data cannot show. On a windy competition day in the Kanto region, I stood beside the 14th green and recorded the ball positions of 12 groups. What I saw was that players with high SG: Approach did not hit the ball lower or harder than others. They chose a different landing point. They aimed at the wide part of the green, accepting distance from the pin, instead of aiming straight at the flag and taking the risk. This is an insight the statistics do not display: target discipline matters as much as technique.

Now to the Vietnam-and-Japan comparison, which I always have to handle carefully because it easily becomes a cultural tic. When I place two groups of young golfers side by side, one trained in the Vietnamese method (emphasising feel, with little measurement) and one in the Japanese method (emphasising repetition, measuring every shot), the difference is not in driving distance. It is in the standard deviation of proximity to the hole.

The Japanese group had an average proximity of 9.8 metres with a standard deviation of 2.1 metres. The Vietnamese group had an average proximity of 10.2 metres with a standard deviation of 3.4 metres. The averages are nearly equal, but the standard deviations differ by more than 60 percent. In other words, the Vietnamese group produced excellent iron shots interspersed with very poor ones, while the Japanese group was more stable. At the top level of golf, stability is worth more than a moment.

But I do not want to turn this into a cheap cultural conclusion. A high standard deviation is not "bad"; it is the consequence of a different coaching philosophy. The point is that at the professional level, when every player is good, the decisive variable is who makes fewer mistakes, not who has the prettier swing. This is something my data, with a small and unrepresentative sample, can only suggest, not assert.

"Gaps in a spreadsheet can also speak, if we are willing to listen." The gap here is this: I do not have GPS training data for the Vietnamese group. I do not know how many hours they train, what they train, or under what pressure. Without that data, any cultural comparison is just a guess dressed up with numbers. And I refuse to do that.

Now I have to argue against myself. The entire argument above rests on an implicit assumption: that SG: Approach measures a stable, repeatable skill. But what if most of the SG: Approach gap between players does not come from skill, but from competition conditions?

This is the biggest blind spot in SG-based golf analysis. The same 150-metre iron shot can produce very different results depending on wind, humidity, green firmness, and time of day. When I split the data by tee-time group (early and late), the average SG: Approach gap between the two groups reached 0.4 strokes per round, purely because course conditions changed. That figure is close to half the gap between the world No. 1 and No. 30.

In other words, a significant part of the "approach skill" I praised may simply be "luck of the tee-time draw". "I do not believe in luck; I believe in cultivated probability." But cultivated probability still needs to be separated from random probability, and I have not yet done that completely.

I tried to control for this variable by comparing only players within the same tee-time group. The result: the correlation between SG: Approach and ranking fell from r = 0.71 to r = 0.63. Still strong, but significantly lower. This is an example of how a contextual variable can inflate a conclusion.

And this is what I want readers to take away: do not trust a metric just because it is printed in bold on a statistics sheet. Ask under what conditions it was measured, how large the sample is, and who chose the metric. In my case, I chose SG: Approach because it is stable and predictive, but I know it contains a noise component I have not fully isolated.

So what is the signal for the next round? If you follow Japanese golf for the rest of the season, watch a metric that is rarely mentioned: proximity from 125 to 175 metres. This is the distance band where most decisive approach shots occur, and also the band where my data shows the clearest difference between players. Track it week by week, and you will see the structure of the season emerge before the leaderboard reflects it.

And the question I still cannot answer: does SG: Approach truly measure skill, or does it only measure conditions? I will return to it when I have more data. For now, I choose to keep the question open, because a conclusion closed too early is usually a wrong one.