Building a Translation Table Between Bay and Turf

An offset is a subtraction. The bay says 158, the table says take six off a seven iron in autumn, so the card says 152 and marks where the correction came from. Simple to apply and genuinely awkward to derive, because you are trying to measure the difference between two environments using a human being as the measuring instrument, and human beings are not stable across four hours.

The Design: Paired, Alternated, Same Morning

The obvious approach is to hit fifty balls indoors, drive to the lane, hit fifty more, and compare the means. We did exactly that for the first two years and the resulting table was wrong in ways that took us a while to see.

The problem is that everything else changed too. The golfer was warmer at the second session. The air was different by mid-afternoon. Whichever environment came second benefited from an hour of extra swings. Any difference between the two sets contains all of that alongside the environment effect, and there is no way afterwards to separate them.

The current design is paired and alternated:

• Two staff, the same two every time, both with stable and well-documented deliveries.
• Five reference clubs, physically the same clubs since 2015, re-measured for loft and lie before each session.
• A single ball model, bought in one batch per year and stored indoors.
• Blocks of eight strikes, alternating bay and lane, over a single morning.
• Air data logged at the start of every block in both places, not once for the day.
• Order reversed between the two staff, so any warm-up drift cancels rather than accumulates.

Alternating is the part that matters. If bay and lane blocks are interleaved, a golfer who is gradually loosening up contributes to both sides roughly equally, and the difference that survives is much closer to the environment effect on its own.

Why the Offset Has to Be Per Club

A single number for the whole bag would be tidy and would be wrong, because the mechanisms behind the offset scale differently with loft and with flight time.

A driver spends six or seven seconds in the air. Air density acts on it for all of that time, and almost the entire flight after the screen is model output. A pitching wedge is up for perhaps four seconds, climbs steeply, and lands steeply, so both the modelled portion and the air's opportunity to act are smaller.

The mat effect scales the other way. It matters most for clubs that interact with the ground aggressively, which is why fairway woods off the deck carry a larger and much more variable offset than their flight time alone would predict.

Put those together and the offset curve across a bag is not a straight line. Ours peaks at the driver, dips slightly at the three wood off a tee, rises again for the three wood off the deck, and then declines steadily through the irons.

Three Ways to Build a Table That Looks Right

These are the traps. We fell into the first two.

Regression to the mean. If you select the reference strikes indoors by quality and take everything outdoors, you have applied a filter to one side only. The indoor sample gets quietly better and the offset grows. Both sides need the same acceptance rule, written down before the morning starts, applied by the same person.

Rollout contamination. Measuring where the ball finishes outdoors and comparing it to modelled carry indoors compares two different quantities. Outdoors we mark the pitch mark, not the resting place. On soft ground that is straightforward; on firm summer turf it needs two people and a spotter, which is why the summer sessions take longer.

One good morning. A single session produces a number with a spread around it wide enough to swallow the effect being measured. Our published offsets are the trimmed mean of at least four sessions per club per season, and if a season only yields two usable mornings, that season's column carries the previous year's figure and says so.

An offset table is a measurement of a room, not a fact about golf. It cannot be borrowed, downloaded, or applied to a different building.

What Goes on the Card

Three things, always. The raw bay reading, the offset applied, and the issue date of the table the offset came from. That way the arithmetic is visible and the correction can be undone by anyone who disagrees with it.

The last row is worth dwelling on. The band gets wider after the correction, not narrower. Applying an offset adds the uncertainty of the offset itself to the uncertainty already in the reading, and a corrected number that claimed to be more precise than the raw one would be arithmetically impossible.

Why Ninety Days

Because Oklahoma does not hold still. Between a February morning and a July afternoon, air density moves enough to shift a driver by several yards. Turf firmness on the lane changes with rainfall, and that changes both rollout and how the ground behaves under a fairway wood. Neither of those drifts slowly enough for an annual table to be honest.

So the table is reissued four times a year and carries a date. If your block happens in the last week before a reissue, the card says which table was used and you are welcome to ask for the updated figures once they exist. Three people have. We sent them.

A table with no date on it is making a claim about permanence that nobody has checked. That is worse than publishing no table at all, because it invites people to trust a correction that may have been derived in a different season under different conditions.

Corrections: none since publication. The paired-and-alternated design described here replaced the older sequential design in 2018; figures published before then are archived rather than deleted.