Measuring the Result — Chronographs and the Statistics of Groups

Measurement is where handloading either becomes an engineering activity or stays a hobby with confident opinions in it. The equipment available to a private bench is genuinely good, and it is also genuinely limited in ways that are rarely stated. Knowing where those limits fall is what separates a conclusion from a coincidence, because the two look identical on a range report.

Volume 21 established the method for developing a load and the reasons most published optimisation methods promise more than they can deliver. This volume is about the instruments those methods depend on: what they measure, how well, how they fail, and what a group on paper does and does not establish.

22.1 Four Ways to Measure a Bullet’s Speed

Optical screen chronographs place two light-sensitive screens a known distance apart and time the shadow passing over each. They are the oldest approach available to a private user and the cheapest. Their weaknesses are all environmental: they want consistent light, they object to direct sun and to deep shade, they need the bullet to pass through a defined window, and they are placed downrange where a misaligned shot can destroy them. They are also the arrangement most likely to be unusable on an indoor range or in bad weather.

Magnetic chronographs — the MagnetoSpeed pattern — strap a bayonet-shaped sensor to the barrel and detect the bullet’s passage magnetically. They cannot be missed and do not care about light, which makes them reliable in conditions that defeat optical screens. The cost is that the unit is attached to the barrel, which changes the barrel’s harmonics. This matters: a group fired with the sensor fitted is not necessarily the group the rifle shoots without it, so velocity and accuracy generally cannot be measured in the same string.

Doppler radar units — the LabRadar pattern and the newer compact units of which the Garmin Xero C1 Pro is the best-known — sit beside the shooter, touch nothing, and track the projectile directly. They remove both the placement problem and the barrel-contact problem, which is why they have displaced the other types for serious work. They have their own failure modes: they need to be aimed, they can lose a shot, some are sensitive to muzzle brakes and suppressors, and the compact units can struggle with small or slow projectiles.

None of them is a pressure gun. Velocity is a proxy. It is the best proxy available at a private bench, and Volume 21’s documented case — a load that produced nearly the predicted velocity while generating enough pressure to shear a bolt handle — is the reason it cannot be treated as more than that.

22.2 The Number That Governs Everything Else

Here is the figure that should shape how every velocity result in this dive is read, and it is not a manufacturer specification.

Two good chronographs, measuring the same shots, agree with each other only to about 7 to 10 feet per second.

That is agreement between instruments, which is the honest way to bound real-world uncertainty — a manufacturer’s claimed accuracy is a laboratory figure and does not include aiming, placement, temperature, or the unit’s own shot-to-shot scatter.

Set that against what handloading does. A tenth of a grain of powder is worth single digits of feet per second. So the quantity a careful handloader controls to a tenth of a grain produces a velocity change at or below the level at which two instruments agree.

Three consequences follow, and they are the practical content of this volume.

A velocity difference of a few feet per second between two loads is not a finding. It is inside the noise.

A “flat spot” in a velocity ladder narrower than about 10 fps is an artefact of the instrument, not a property of the rifle. This is the cleanest form of the criticism in Volume 21, and it does not require any statistics to state.

Comparisons made on one chronograph on one day are more trustworthy than comparisons across instruments or across sessions. The unit’s systematic error largely cancels within a session and does not cancel between them.

22.3 Standard Deviation, Extreme Spread, and Which One to Report

Two statistics dominate range reports, and one of them is much worse than its popularity suggests.

Extreme spread is the difference between the fastest and slowest shot in a string. It uses exactly two shots and discards every other shot fired. In a five-shot string it is built from 40 percent of the data; in a ten-shot string, 20 percent. It is also guaranteed to grow with sample size — a twenty-shot string will almost always show a larger extreme spread than a five-shot string from the identical load, which means extreme spreads from different sample sizes cannot be compared at all. That property alone disqualifies it as the headline number, and it is the number most commonly quoted.

Standard deviation uses every shot. It is the better statistic and it comes with its own caveat: a standard deviation computed from five shots is itself a very noisy estimate, and quoting it to a tenth without the sample size beside it implies a precision that is not there.

The workable practice is to report standard deviation, always with the sample size, and to treat extreme spread as a rough indicator of whether something went badly wrong rather than as a measure of quality.

22.4 The Statistics of Group Size

Group measurement is where small samples do the most damage, because a group is a physical object that looks like evidence.

A five-shot group has roughly plus or minus fifty percent natural variation in its size. Two five-shot groups fired from the identical load, from the same rifle, in the same conditions, routinely differ by that much. A load that produced a group half an inch across and one that produced three-quarters of an inch may well be the same load.

A three-shot group tells you almost nothing. Three shots is the smallest number that produces a two-dimensional shape, which is precisely why it is seductive — it looks like a result. Its variation is larger still, and the practice of firing several three-shot groups and reporting the best one is not measurement; it is selecting the most flattering sample from a noisy distribution.

Separating two genuinely different loads takes on the order of fifty rounds or more to a reasonable confidence. That figure is the honest cost of the question “which of these two loads is more accurate”, and it is why Volume 21 recommends spending components on fewer questions rather than more.

Extreme spread has the same defect on paper as it does in velocity. The conventional group measurement — the widest distance between two holes — uses two shots out of however many were fired. Mean radius, the average distance of each shot from the group’s centre, uses all of them and is the better statistic for the same reason standard deviation is. It is less familiar and harder to measure by hand, which is the only real argument against it.

22.5 Making a Group Measurement Mean Something

Several practices cost nothing and materially improve what a range session establishes.

Fire enough rounds to answer the question being asked, and if there are not enough components for that, ask a smaller question.

Alternate between loads rather than firing all of one and then all of the other. Firing in blocks confounds the comparison with barrel heating, changing light, changing wind and the shooter settling in. Alternating spreads those across both candidates.

Record the conditions, because temperature is not a nuisance variable in handloading — powders differ in how much velocity they gain per degree, and a load developed in summer and verified in winter has been tested twice for a reason.

Separate the two measurements. Velocity work with a barrel-mounted sensor and accuracy work cannot honestly be done in the same string, for the harmonics reason above.

Shoot from something that removes the shooter. A group is a measurement of rifle, ammunition and shooter together, and the shooter is usually the largest and least stable of the three.

22.6 What the Instruments Cannot Tell You

Two limits are worth stating plainly, because a well-equipped bench can otherwise produce a false sense of completeness.

Nothing here measures pressure. Not velocity, not case-head expansion, not primer appearance. Volume 21 covers why the popular pressure signs are weak and which ones carry weight; the honest position is that a private bench works inside published data and uses signs as a stop condition, not as a measurement.

A chronograph cannot validate a simulation. Pressure-prediction software that agrees with a chronograph on velocity has been confirmed on the output the user can check and not on the one the user cannot. Volume 21 documents the case where exactly that happened.

22.7 Feeding It Forward

A measurement that is not written down against the specification that produced it is an anecdote. Volume 23 covers the record — what a load record has to carry, how batches stay traceable, and how a range result becomes an input to the next batch rather than a memory of a good afternoon.