Measuring Groups: Why Extreme Spread Throws Away Most of Your Data
You shoot a ten-shot group, put the calipers on the two widest holes, subtract a bullet diameter and write down 0.84 inches. That number is how essentially everyone reports accuracy — and it was calculated from two of your ten shots. The other eight, containing most of the information you paid for in powder and barrel life, were ignored entirely.
Worse, the two you used are the least representative ones you had. Extreme spread is measured from your two worst outliers, which means it is a measurement of your flyers, not your rifle.
Why extreme spread behaves badly
Because it is defined by the two most extreme shots, extreme spread can only ever get larger as you shoot more. Fire five more rounds into the same group and the number cannot improve — it can only stay the same or grow. That makes it useless for comparing groups of different sizes, which is exactly what people use it for.
It is also wildly noisy. Two groups from the same rifle and the same load can differ by a factor of two on extreme spread purely by chance, because you are sampling the tail of a distribution with a handful of shots. This is why load development based on “which charge shot the smallest group” so often fails to reproduce: you were reading noise and calling it signal.
Applied Ballistics published conversion factors that show the scale of the problem — the same rifle measured by extreme spread reads progressively worse simply because you shot more:
| Comparison | Multiply by |
|---|---|
| 3-shot → 5-shot | × 1.28 |
| 3-shot → 10-shot | × 1.58 |
| 3-shot → 20-shot | × 1.85 |
So a rifle honestly reporting 0.5 MOA on three-shot groups is the same rifle that reports 0.79 MOA on ten-shot groups. Neither number is a lie. But quoting the three-shot figure and comparing it against someone else’s ten-shot figure is meaningless, and it happens constantly.
Mean radius: use all the shots
Mean radius is the average distance of every impact from the group’s centre. Find the centroid, measure each hole’s distance from it, take the average.
Every shot contributes. One unlucky flyer moves the number a little rather than defining it entirely. And because it is an average rather than an extreme, it stabilises as you add shots instead of drifting — more data makes the answer better, which is how a measurement ought to behave.
Bryan Litz makes the case plainly: in a ten-shot group, extreme spread uses information from 20% of your shots. Mean radius uses 100%.
The practical difference
Where this bites hardest is comparing two loads. Load A shoots nine tight holes and one flyer; load B shoots ten evenly mediocre ones. Extreme spread may well rank load B as better, because A’s flyer defines its whole score. Mean radius correctly identifies A as the more accurate load with one bad round in it — which is almost certainly the truth, and almost certainly the load you want.
How to measure either one properly
- Extreme spread is measured outside-to-outside between the two widest holes, then minus one bullet diameter to get centre-to-centre. Forgetting the subtraction is the single most common measurement error, and it inflates a .308 group by nearly a third of an inch.
- Convert to angular units so distances are comparable:
MOA = inches ÷ (1.047 × yards/100)andmils = inches ÷ (3.600 × yards/100). Note the 1.047 — true MOA, not the rounded inch. - Record the shot count with every group. A group size without its n is not a measurement, it is a boast. “0.6 MOA” means nothing until you know whether that was three shots or twenty.
- Measure from a known distance, lasered rather than assumed.
Print the group target
Four bulls with half-MOA rings and per-group data blocks, a caliber subtraction table, and both conversion formulas printed on the sheet. The one-inch calibration bar doubles as the scale reference if you photograph the target to measure it.
Group Analysis Target (PDF)What it costs you to get this wrong
Mostly, you make decisions from noise.
A load-development session judged on the smallest three-shot extreme spread picks a charge weight that won a lottery. You load a hundred rounds on it and it does not reproduce, so you conclude the rifle is inconsistent, or the brass needs annealing, or the barrel is going. The actual problem was that you compared seven very small samples using the metric most sensitive to outliers, and the winner was chance.
The same error corrupts everything downstream. A barrel gets replaced because its groups “opened up” when the shooter simply moved from three-shot to five-shot groups. Components get blamed for what was really a sampling artifact. Money and barrel life go into chasing a difference that was never there.
And it distorts your ballistic data too. If you cannot separate genuine dispersion from a couple of bad shots, you cannot tell whether a miss at distance was your load, your scope or you — which means you cannot true your solver with any confidence, because you do not know which of your impacts deserve to be trusted as data.
What it changes for you
You start being able to tell real differences from imaginary ones. Two loads that genuinely differ will separate on mean radius across a reasonable sample; two that only differed by luck will not. That alone saves components, barrel life and a great deal of second-guessing.
It also makes your own records comparable over time. Group data recorded with its shot count, distance and measurement method is data you can still use in two years — to see whether a barrel is actually degrading, whether a new lot of bullets changed anything, whether that load really was better. Group data without those things is a photo album.
The habit worth building is small: log every group with n, distance, and both numbers if you have them. It costs nothing at the bench and it is the difference between a shooting record and a shooting scrapbook.
Next: ladder test vs OCW — two methods people constantly confuse