Why We Pool by Speed Band Instead of Averaging Everyone Together

The first version of the offset table, back when Teddy and I were still arguing about how to present it, was a single figure per club. One number for driver, one for six iron, one for pitching wedge. It was easy to print and it was quietly wrong for most of the golfers who read it.

What a Flat Average Does to a Real Pattern

Take the driver. Across every paired block we have ever run, the bay overstates carry, and it overstates it more as clubhead speed increases. A golfer swinging under 88 miles an hour sees the bay run a little over two yards long. Someone above 106 sees it run more than seven yards long. Average those two golfers together and you get a middle number that describes neither of them accurately. The slower player gets handed a correction that is too aggressive. The faster player gets handed one that is not aggressive enough.

This is not a subtle effect once you plot it. The gap grows because the flight model is extrapolating from a short indoor look at the ball, and there is simply more flight left to predict for a fast, low-spinning shot than for a slower one. Collapsing that relationship into a single average erases the exact thing a correction table exists to capture.

Why We Draw the Bands Where We Draw Them

The four driver bands on our card — under 88, 88 to 97, 98 to 106, and above 106 — are not round numbers chosen for tidiness. They come from where the offset actually bends in the paired data. We tried finer bands first, six or seven of them, and found the sample inside each one got too thin to trust. We tried two bands and found the middle of the range got smeared across a boundary that did not actually exist in the data. Four is where the bands stopped changing meaningfully as we added more seasons of pairs.

Irons behave differently, which is part of why this matters. A six iron reads only about two yards long, and that figure barely moves across speed. Splitting the iron table into speed bands the way we split the driver table would just be adding noise to a number that is already stable. So the iron rows on the card sit by club alone, and the driver and fairway wood rows split by speed, because the underlying pattern in each category is different and the table should reflect that rather than force every club into the same format for consistency's sake.

What Gets Lost If You Pool Too Coarsely

The clearest case I have seen was a golfer in the 88 to 97 band who had been handed a flat, unbanded correction from a printout at another shop. His actual bay-to-turf gap for his own speed sat close to four yards. The flat correction he had been given, averaged across a much wider population, was closer to five and a half. He had been building his approach yardages off a number that was overcorrecting by a club and a half of carry on some par threes. Nothing about his swing was the problem. The table he was handed had smoothed away the information that would have told him where he actually stood.

What Gets Lost If You Split Too Finely

The opposite failure is just as real and less obvious. Splitting a category into more bands than the sample supports produces rows that look precise and are actually just noisy. If a band only has six or seven pairs behind it, the number on the card is closer to a guess wearing a decimal point than a measurement. That is why the confidence column exists on the iron table and why we will say “fewer pairs, moderate confidence” rather than print a number we cannot stand behind.

The Rule We Actually Follow

Split the data as far as the pattern genuinely changes and the sample genuinely supports it. Stop before that. It is a boring rule and it produces a table with an uneven number of rows per club, which looks less tidy than a single figure per category. It is also the only version of the table that tells you something true about your own speed rather than about an average golfer who does not exist.

Nate Brogan runs the driver and iron benches at Dialed In Lab in Eugene, Oregon.