Why Strokes Gained Putting Flatters and Fools Club Golfers

Every winter, a few thousand UK club golfers look at their tracking app, see a positive strokes gained putting figure, and conclude that the putter is the one part of the bag that’s working. They then spend February on the indoor mat grooving a swing change, because “the putting’s fine.”

The putting is probably not fine. It’s probably invisible. What that number is actually showing you, most of the time, is how close your approach shots finished.

The per-putt maths barely moves

Strokes gained putting is meant to be distance-adjusted, and that’s exactly why people trust it. You get compared against an expected-putts baseline for the length you actually faced, so a 40-footer and a 6-footer are judged on their own terms. In theory, nothing about proximity leaks in.

In practice, run the expected value for a real club golfer and the adjustment cancels itself out. Take a 12-handicapper with typical conversion rates, compared against a PGA Tour baseline:

Look at the right-hand column. It sits between −0.05 and −0.14 across a ten-fold range of distance. That’s the system working as designed: on average, this golfer loses about a tenth of a shot per putt whether the ball is four feet away or forty.

Which leaves a problem. If the expectation is flat, then everything you observe in a single round is noise around that flat line, and the size of the noise is set by how many live make-or-miss situations you were handed. Two-putting from 40 feet is worth +0.06. Two-putting from 20 feet is worth −0.13. Neither one moves your round. The needle only moves when a putt drops from somewhere it plausibly might not have, and that window is roughly six to twenty-five feet.

You do not get into that window with your putter. You get there with a 9-iron.

The ceiling is set by your approach play

Imagine the best long-putting round of your life. Eighteen greens hit, every first putt between 35 and 50 feet, and you two-putt all eighteen. No three-putts at all, which no club golfer has ever done. Against the tour baseline that’s about +1.5 for the round, and against a handicap-matched baseline it’s closer to +1.8, because peer golfers three-putt from that range perhaps 15 to 20% of the time.

Now imagine an ordinary putting day where you happened to strike it beautifully. Twelve greens, average proximity 21 feet, and four putts drop from 9, 12, 16 and 22 feet. Against a 12-handicap baseline those four makes are worth +0.69, +0.79, +0.88 and +0.96 respectively: +3.32 from four swings of the putter. Give back 0.78 across thirteen neutral two-putts and 0.89 on a three-putt from 34 feet, and you’ve still posted +1.65 for the round while taking 33 putts.

Thirty-three putts. Best putting round of the season, according to strokes gained. The same player, same stroke, same read quality, on a day when his average first putt was 38 feet, comes in around +0.8 while taking 36 putts.

That’s a 0.85-stroke swing in the putting statistic driven entirely by where the ball started. Nothing in the stroke changed. This is the whole trick: proximity manufactures the opportunities, and strokes gained putting counts the opportunities you converted without ever telling you who manufactured them.

Your chipping is also hiding in there

Miss eight greens, chip four of them to inside five feet, hole all four. Each of those conversions is worth roughly +0.2 against a peer baseline, because club golfers miss from that range about 18% of the time. That’s +0.8 of “putting” that is really wedge quality with a different label on it.

Scramble badly on the same eight holes and you’re facing 14-footers instead, converting maybe two of them, posting a similar number through a completely different mechanism. The stat cannot distinguish. It only sees the distance of the first putt on the green, and it credits whoever holed it.

Half of you are measuring the input wrong anyway

None of the above matters if the distance going into the model is fiction, and for a large share of club golfers it is.

Shot Scope’s V5 and X5 make you press the watch on the green to log the actual pin position, and that single button press is why their putting numbers are worth reading. Arccos Caddie, with Smart Sensors at around £180, auto-detects shots brilliantly and claims detection accuracy in the mid-nineties for full swings, but putts are where users edit most often, and if you never set the pin, the system works from a computed position rather than the flag you were aiming at. On a 30-yard-deep green, a centre-of-green default can be 15 yards wrong. That turns a genuine 12-foot putt into a recorded 45-footer, and hands you free strokes gained you did not earn.

TheGrint, Golfshot and V1 Game all derive strokes gained from distances you type in by hand, which is honest but only as good as your eye. Clippd aggregates whatever you feed it into a Player Quality score, and inherits every error underneath. If you want the wider framework for what these systems can and cannot tell you across the bag, the pillar on strokes gained and shot tracking for amateurs covers how the categories interact.

What phone video and AI can actually read on a putting stroke

Here’s where a lot of money gets wasted. Face angle at impact controls roughly 80 to 85% of a putt’s start line. To hole a straight ten-footer you need start-line control inside about 0.6 degrees, because the hole is 4.25 inches wide and the ball is 1.68 inches, leaving a little over an inch of margin either side at 120 inches of travel.

Now consider what your phone is capable of. Shooting 4K at 60fps gives you 16.7 milliseconds between frames; a putter head moving at 1.4 m/s travels about 23mm in that gap, and the head is motion-blurred across the frame. Drop to 1080p at 240fps and you’re down to roughly 5.8mm of travel per frame, better, but phone sensors use rolling shutter, reading the image line by line over 10 to 30 milliseconds. A moving putter head comes out geometrically skewed. That skew looks exactly like a face angle change to any vision model trying to read one.

No amount of AI resolves this, because the information is not in the file. Any app promising face-angle-at-impact feedback from handheld phone footage is generating a plausible-looking number from a blurred, skewed, sub-degree signal. Treat it as decoration.

What genuinely measures putting face and roll: Quintic Ball Roll, using high-speed cameras at 360 to 1080fps in a controlled setup, which is a fitter’s system rather than a garage purchase. SAM PuttLab does it with ultrasound across 28 parameters, at several thousand pounds. HackMotion, at roughly £300 to £500, straps to your lead wrist and samples flexion and extension at high rate; it is a direct measurement, not an inference from pixels, and wrist angle through the stroke is tightly coupled to face control. Neither a Garmin Approach R10 nor a Rapsodo MLM2PRO will give you usable putting data, whatever the marketing implies about their launch monitor credentials.

Phone video is excellent for the slow, large stuff. Eye position over the ball, shaft lean at address, ball position relative to your lead shoulder, head movement during the stroke, backstroke-to-through-stroke length symmetry, and tempo ratio (many tour putters sit near 2:1 back-to-through in time). Onform, V1 Golf and Hudl Technique all handle frame-by-frame comparison and drawing properly and don’t pretend to be measurement systems. Sportsbox 3D Golf reconstructs markerless 3D body kinematics from a single phone camera and is credible for torso and pelvis motion, which is useful for full swings and fine for putting posture, but its resolution was never designed for a half-degree clubface.

So what do you actually practise

Speed control is where the recoverable strokes are. Ten balls from 40 feet, count how many finish inside a three-foot radius; a good club golfer manages five or six, tour players are near eight. Every one that finishes outside that circle is a live three-putt, and three-putts are the one putting outcome that shows up cleanly in the data rather than drowning in noise.

Then build a make-rate log from eight feet. Thirty balls, gate drill, write the number down over six sessions. A 12-handicap typically sits around 30 to 38% there; scratch is nearer 45 to 50%. That’s a real trend line, unlike your round-to-round strokes gained figure, which will bounce by two full shots on stroke quality that hasn’t changed at all.

And then give the rest of your practice hour to wedges from 50 to 110 yards. Tour average proximity from 100 yards is about 18 to 20 feet. A 15-handicapper from the same spot averages north of 45 feet and misses the green something like 35 to 40% of the time. Broadie’s data puts approach play at roughly 40% of the gap between a 90-shooter and a tour professional, against about 15% for putting. Close the proximity gap by ten feet per approach and your putting statistic will improve without you touching the putter, because you will finally have putts worth holing.

Try this over your next five rounds: log the pin position every time, and record approach proximity alongside first-putt distance. Put the two columns side by side. The correlation you find will tell you more about your golf than the putter you were about to buy.