Hideki Matsuyama's Adjusted Strokes Gained at the 2026 Sony Open: When the Two-Metre Putt Zone Says What the Scorecard Hides

On the 14th hole at Waialae Country Club, Hideki Matsuyama's putt measured 2.1 metres. The ball stopped 12 centimetres from the cup. It was the fourth miss from inside 2.5 metres in seven consecutive holes. The scorecard simply recorded "Par" — a neutral number that reflects nothing unusual. But when I pulled the ShotLink data from that round and cross-referenced it against our internal Strokes Gained: Putting model, the gap between actual score and expected score reached -3.4 strokes in holes 13–18 alone.

This is the kind of data the traditional scorecard conceals. And it forced me to re-ask a question I had not thought I would have to ask all season: does the putting curve of a 33-year-old major winner still follow the age model I have always trusted?

To contextualise this number, we need to return to the structure of the model. The Strokes Gained: Putting (SG: Putting) I use in my analysis is not the official PGA Tour metric — I recalibrate it using actual putting-distance data from ShotLink, because the original model buckets distances in 3-foot bands, whereas I subdivide by 0.5 metres to capture deviation in the 2–3 metre zone.

Why does the 2–3 metre zone matter? Probabilistically, it is a transition zone — where tour professionals' success rates range from 42% to 58% depending on the individual, and where the best putters generate their edge. Over the past three seasons, Matsuyama's SG: Putting from 2–3 metres sat at +0.18 strokes per round — meaning he scored better than the tour-average expectation in this zone.

At the 2026 Sony Open, that number fell to -1.92 strokes per round. This is not random variance. With a sample size of 72 holes, the standard deviation of this metric for Matsuyama over the past five seasons is 0.4. A reading of -1.92 falls outside the 99% confidence interval.

The more important context: this was the third tournament since the shoulder injury he sustained in November last year. And it is the first time in my 17 years of tracking golf data that I have observed a post-injury golfer whose putting metrics declined more sharply than their long-game metrics.

The evidence chain starts here. I took all four rounds of Matsuyama's 2026 Sony Open and split them into six-hole blocks for comparison:

Holes 1–6: SG: Off the Tee +0.82; SG: Approach +0.44; SG: Around the Green +0.21; SG: Putting -0.15; Total SG +1.32.

Holes 7–12: SG: Off the Tee +0.71; SG: Approach +0.38; SG: Around the Green +0.12; SG: Putting -0.48; Total SG +0.73.

Holes 13–18: SG: Off the Tee +0.65; SG: Approach +0.29; SG: Around the Green -0.31; SG: Putting -1.18; Total SG -0.55.

What does this table show? Matsuyama's long game does not decline significantly — SG: Off the Tee and SG: Approach hold steady. But there is a distinct shift over time: SG: Putting collapses in the late-hole group, dragging SG: Around the Green down with it. Matsuyama's performance is a problem of the closing stretch, not of the entire round.

I dug deeper into the ShotLink data to test the assumption. There are three hypotheses for this decline: physical fatigue late in the round; psychological focus decay; or a specific technical fault in short putting.

Testing the fatigue hypothesis: I cross-referenced Matsuyama's GPS movement data. In Round 3, he covered 8,432 metres — below his season average of 8,750 metres. No evidence of physical fatigue beyond threshold.

Testing the focus hypothesis: this is harder to rule out directly, but there is a proxy. If focus were the issue, I would expect to see uniform decline across all putting distances. In reality, his SG: Putting from 1–2 metres declined only -0.4, while from 2–3 metres it declined by -1.52. That asymmetry does not match a focus-decay model.

Testing the technical hypothesis: this is where my assumption collapses. I reviewed the video of 12 putts from 2–3 metres. Nine of them shared the same characteristic: the putter face opened 2–3 degrees at impact. This is not a fatigue issue — it is a specific mechanical issue in the left wrist, potentially related to the November shoulder injury.

I asked the wrong question. I was looking for a time-based decline pattern, when the data pointed to a specific mechanical error occurring randomly. The gap in the spreadsheet is not "when did he putt poorly" — it is "why does the same error appear on some putts and not others".

If I concluded that the shoulder injury was the sole cause, I would commit the classic correlation-causation error. There are at least two other factors I have not yet entered into the model: the green conditions at Waialae, and the schedule.

Waialae has Bermuda greens running at Stimp 11.5 — faster than the tour average in Hawaii. A golfer with a grip adjusted after injury may respond to fast greens differently from a normal golfer. This is an untested hypothesis because I do not have grip-pressure data.

Second, Matsuyama has played three consecutive weeks since the Sony Open. That schedule is unusual for a golfer recovering from injury. It is possible the mechanical fault stems not from the shoulder but from adjusting too quickly to a new grip.

And here is where I must criticise myself: my model lacks a real-time grip-pressure variable. Over 17 years of analysis — from my first mistake at J.League 2026 to the PPDA lesson at the 2026 World Cup — I have learned that data is only trustworthy after contextualisation. But I have still not learned how to handle it when the very variable I need lies beyond my capacity to collect.

Every number is a confession not yet written into prose. And this confession is: I built a model that can measure putting, but cannot measure what lies between the hand and the ground.

The signal for the next tournament is not "Matsuyama is putting poorly" — that conclusion is too broad. The signal is: if his SG: Putting from 2–3 metres continues below -1.0 at the Farmers Insurance Open, and a grip-pressure fault emerges, then this is a mechanical problem rather than a psychological one. Conversely, if the metric recovers to around +0.1, it is likely just variance across three post-injury rounds.

I still do not have the data to answer the original question. And that means it is I — not Matsuyama — who needs to adjust the model. What did NOT happen on the remaining putts often speaks more truth than what did happen on the missed ones.