Right, the data points are there but which data points are more accurate than the others?
How to identify and weed out the bad data points?
One unit may record lower ES or SD than the other but which one is correct?
The spread between all 6 units for 80% of the shots was less than 4.2fps. A 4.2fps error in a ballistic engine for my 6 creed load would be a MAXIMUM potential of 0.84" at 1k and with a simple RSS assessment on a rifle shooting .5moa at 100yrds shows it would only account for less than .1" at 1,000yrds. Equally, realize, the average displayed by all 6 units were only 2.7fps spread from fastest to slowest, so HALF of that potential difference at 1,000yrds. Within this dataset, ALL of the singular results were incorrect, because the variability of the ammo itself, but for each data point, the potential error from one unit to the other was very tight. So as I mentioned above, either ALL of them are good enough, or none of them are. A 99% confidence interval for this particular dataset, however, shows any given one of these units would only be potentially within ~9fps of the true average - even after 100rnds - so this dataset isn't seeking truth, it's seeking comparison. When we're only an average of 3.5fps spread among the 6 units with a set which only held +/-9fps for a 99% confidence interval, eh, nah, they're either all telling the truth, or none are. Even if I only took ONE shot, the odds of being "wrong" have nothing to do with the unit being used to measure the velocity, and everything to do with the consistency of the ammunition and the relative insensitivity of our trajectory to such small variations in velocity.
Vertical spread is a half decent litmus test for validating recorded speeds and figuring out which unit is lying. Barrel whip not-withstanding.
I'll readily admit, I'm not sufficiently skilled as a shooter to be able to tell the difference of 0.06" at 1,000 yards in my groups to determine truth between two chronographs showing 4fps spread among them. Certainly, my subsequent testing using a better performing centerfire load instead of this cheap and shitty 22LR ammo will improve ability to determine validity of result, but again, I'd have to be able to hold less than 1/10th of an inch difference at 1,000yrds to be able to tell the difference.
Anecdotally, I've used my Garmin velocity over the last 2 seasons to hit 1.5moa targets with 1st round impacts as far as 2200yrds. This is with a load which typically displays 5-9fps SD's for 10 shot strings, and 7-10 SD's for 60 shot SD's. A 99% Confidence interval for this load would be +/-3.2fps. That potential 6.4fps variability in the 1% uncertainty describes a difference of 19.16mils vs. 19.25mils at 2200yrds... One click in my scope, on a target about 5 clicks tall - and I'll be honest, I just don't shoot well enough to hold my shots within 1 click on the turret at 2200yrds.
So I struggle a little to conceive of any application where I REALLY have the ability to differentiate with live fire between any results as close together as the above dataset describes. I generally have control at 2000-2200 where my waterline is within ~0.3mils using velocity obtained via my Garmin, so to think I'd dial off 0.02mil less if I were using the LabRadar or dial on 0.03mil more if I were using the Athlon, eh, I'm pretty sure I can't shoot the difference between the 3 brands.
I do, however, have a MacDonald TwoBox acoustic chronograph unit which claims to be true within +/-1-2fps, which is tighter than the potential +/-0.1% of these radar units. Set up is a massive pain in the ass, requiring level and straight installation across 15ft span, so if this works as described, this would be my most accurate opportunity for "true" velocity within subsequent testing out of the 8 different chronograph units I have on hand. Data for this forthcoming with sunny weather. But reminding, as we see here, we have high confidence that the "truth" is somewhere between the min and max of these 6 units, and despite being the minority result, any of the 6 units each demonstrated the highest, lowest, and median results throughout this dataset.
.
