I found this to be interesting and an official WTH. I'm not going to argue with the guy because I can barely avenger (1+1)/2 www.youtube.com/watch?v=JSxr9AHER_s The video long and applies to PB's and I'd guess it also applies to pellet guns.
Are 5 shot groups worthless statistically?
- Thread starter HogKiller
- Start date
Wonder why 5 shot groups are used so often in BR competition? Better than 3 shot groups and not as wasteful as 10 shot groups. I guess "to each his own".
Thissssss!
Some guns I KNOW I have to put a few or more pellets through to see of it likes em...some I can put just one or even 2 in and go NOPE she ain't having those
Upvote 0
The pump for my Crown doesn't mind 30 shot groups, but the pump for my Sheridan likes five shot groups. 
Upvote 0
SHOTS: ABCDE
PAIRS: AB AC AD AE BC BD BE CD CE DE
It is a statistically significant sample set.
There are 10 data points in the pairs. The average distance between the pairs divided by two represents the average radius of the group.
If you then take the standard deviation of that sample set you can state with a certain confidence the actual deviation of that radius.
The argument the video makes is about the level of accuracy achieved with a five shot group.
He is saying it is very difficult to measure small variations with a small sample set.
Everybody already knows that. It just took him 10 minutes to say it.
PAIRS: AB AC AD AE BC BD BE CD CE DE
It is a statistically significant sample set.
There are 10 data points in the pairs. The average distance between the pairs divided by two represents the average radius of the group.
If you then take the standard deviation of that sample set you can state with a certain confidence the actual deviation of that radius.
The argument the video makes is about the level of accuracy achieved with a five shot group.
He is saying it is very difficult to measure small variations with a small sample set.
Everybody already knows that. It just took him 10 minutes to say it.
Last edited:
Upvote 0
I have read for a while that a 7-shot group is the minimum for statistical reliability. Then again, probably not going to get that many shots off at an animal.
Upvote 0
Ugh, another joker who takes a useful statistical measure and puts an overly broad, overreaching title on the content to make clicks.
Science is only “right” when it is applied to suit to the need.
Reasonable group size is always a trade off of cost/consumables and certainty.
In testing rimfire and airgun, in which the ammo is commercially produced, I tend to use 10 shot groups when I want a bit more certainty. Multiple ten shot groups when I have similarly performing candidates and I’m trying to tease out which is better. If conditions are tricky I will back it down to multiple five or three shot groups to account for condition changes. Ten (or twenty or three) shot groups lose all value if you miss a condition change.
If I’m testing in good conditions to find a good batch or lot of ammo, five shots can be enough to confirm it’s not a good lot for my gun. Once you see a flier, stop shooting that ammo because it will only get worse.
Shooting groups to prove consistency requires the highest group counts. Here you’re hoping to find that 1 in 100 bad round. I attended a “6400” smallbore match last weeekend and 1) I was amazed at the shooting and 2) amazed at the consistency of their ammo. The winner posted a 6399.586, which means he missed the ten ring only once in 640 shots, and 586 shots were straight up bullseyes. Proving ammo for that kind of performance would take a brick just to have an idea of it would hold up for the match. This matters because one competitor in this match that I know once bought two cases of premium match rimfire ammo and found one flier shot per brick. Unusable in such a competition.
The guy in the video know enough about statistics to be sophomoric but not enough to use statistics in the real world. Shooting is not an easy topic to apply statistics usefully. Any statistician who is unwilling to adjust their math and fully exposes the experiment to other bigger variables: wind, production lot size, barrel fouling, barrel wear, etc is conducting a bad experiment. To use his terms: “zero statistical significance”. Statistical certainty on paper while letting those other variables run rampant is worthless.
As for me, I test airgun pellets in observance of the lot quantity. If I have a tin of 350 pellets, I might test 30-50 indoors on a bench to see how good it might be, which leaves 90% for intended purpose. (This guy would claim I didn’t learn anything until I’ve shot ten tins). Outdoors, even with wind flags, I would rattle off a few five shot groups, trying to minimize condition changes and using no hold-off.
Science is only “right” when it is applied to suit to the need.
Reasonable group size is always a trade off of cost/consumables and certainty.
In testing rimfire and airgun, in which the ammo is commercially produced, I tend to use 10 shot groups when I want a bit more certainty. Multiple ten shot groups when I have similarly performing candidates and I’m trying to tease out which is better. If conditions are tricky I will back it down to multiple five or three shot groups to account for condition changes. Ten (or twenty or three) shot groups lose all value if you miss a condition change.
If I’m testing in good conditions to find a good batch or lot of ammo, five shots can be enough to confirm it’s not a good lot for my gun. Once you see a flier, stop shooting that ammo because it will only get worse.
Shooting groups to prove consistency requires the highest group counts. Here you’re hoping to find that 1 in 100 bad round. I attended a “6400” smallbore match last weeekend and 1) I was amazed at the shooting and 2) amazed at the consistency of their ammo. The winner posted a 6399.586, which means he missed the ten ring only once in 640 shots, and 586 shots were straight up bullseyes. Proving ammo for that kind of performance would take a brick just to have an idea of it would hold up for the match. This matters because one competitor in this match that I know once bought two cases of premium match rimfire ammo and found one flier shot per brick. Unusable in such a competition.
The guy in the video know enough about statistics to be sophomoric but not enough to use statistics in the real world. Shooting is not an easy topic to apply statistics usefully. Any statistician who is unwilling to adjust their math and fully exposes the experiment to other bigger variables: wind, production lot size, barrel fouling, barrel wear, etc is conducting a bad experiment. To use his terms: “zero statistical significance”. Statistical certainty on paper while letting those other variables run rampant is worthless.
As for me, I test airgun pellets in observance of the lot quantity. If I have a tin of 350 pellets, I might test 30-50 indoors on a bench to see how good it might be, which leaves 90% for intended purpose. (This guy would claim I didn’t learn anything until I’ve shot ten tins). Outdoors, even with wind flags, I would rattle off a few five shot groups, trying to minimize condition changes and using no hold-off.
Last edited:
I typically do 1 mag or more groups, as in 2 mags, 3 mags, 4 or even 5 sometimes. Sometimes when the 5 , 6 or 7 shots are phenomenally placed I cheat and will stop the group short.
Upvote 0
Why 5 shot groups? Why not? You could compete with only one shot at a small target, or 20 shot groups. As long as everyone has the same challenge, one can argue that it's fair. Obviously, one or twenty shots presents issues of both random luck, and the effects of conditions. I think, over time, group competition based on 5 shots has proven to be a fair and reasonable basis of competition. We could make baseball competition based on two or twenty innings, but a nine inning game has proven to be a reasonable competition. Do we need to search for statistical support? I think not. It's a game, enjoy.
Upvote 0
I look at it this way - the first shot is just that, a first shot, the second shot is a shooter trying to improve on what the first shot’s POI told him, the third shot in my opinion is the only one that matters because the shooter and the gun have come to an agreement of sorts “yeah that’ll work”.
Upvote 0
I found this answer very truthful and the later part honest and hilarious lol. I think we all have pulled that at some time or another.
Upvote 0
There is a corollary: the less you care about group size, the better you will shoot the entire group.I found this answer very truthful and the later part honest and hilarious lol. I think we all have pulled that at some time or another.
When I give myself the mindset that the only group that would raise my pulse is an impossibly small one hole group, it gets easier to not toss a flier. It doesn’t fix it, but it helps.
David
Last edited:
Upvote 0
It is the truth, I am a good shot and I truly believe that, there are some PHENOMENAL shots here on AGN though. I however feel that 5 shots is to easy to repeat, I think 10 shots or more is a true indication as to whether or not a rifle can show us true repeatable accuracy. Thats my opinion though, but really, sometimes I shoot 5 shots at 75 plus and I am just so chuffed I call it there and start a new group lmaoI found this answer very truthful and the later part honest and hilarious lol. I think we all have pulled that at some time or another.
Upvote 0
As many here, I am a researcher. The sample of any size having utility is based on the sampling methodology.
- Is is a truly representative sample of the whole population?
This depends on a few things, not the lease is the variation. How "even" is the variation in the population? Is it normally distributed or not? Is it skewed or blotchy (multimodal)? While central limit theory suggests 30 samples will mitigate some of this (samples with replacement) we can't do replacement here.
Another is that variation is caused by something or multiple somethings. If multiple somethings, how are these known to interact? Often a stratified or propensity sampling process is indicated. Are these mixtures the same in each "batch"? Essentially our pellet sorting is an after-the-fact attempt to address these variation issues.
Then there is also the difference between statistical and practical significance. This is another can of worms. We are after the latter not former.
Further we are not really looking at sample of the batch here. We are usually asking a predictive question. What will the next shot do? So sampling has an added dimension here.
So it is not quite straight forward.
We as "researchers" in shooting pellets are at the least sophisticated research level here. We are not really into the realm of "statistical significance". But, the real research question is ... does it have practical value?
If I find I can drop 5 shots under 1 inch at 50-yards 4 of 5 times using pellet "x" it has practical value to me. I have a practical experience that suggests what will happen next. It may not be statically significant but does have practical significance.
- Is is a truly representative sample of the whole population?
This depends on a few things, not the lease is the variation. How "even" is the variation in the population? Is it normally distributed or not? Is it skewed or blotchy (multimodal)? While central limit theory suggests 30 samples will mitigate some of this (samples with replacement) we can't do replacement here.
Another is that variation is caused by something or multiple somethings. If multiple somethings, how are these known to interact? Often a stratified or propensity sampling process is indicated. Are these mixtures the same in each "batch"? Essentially our pellet sorting is an after-the-fact attempt to address these variation issues.
Then there is also the difference between statistical and practical significance. This is another can of worms. We are after the latter not former.
Further we are not really looking at sample of the batch here. We are usually asking a predictive question. What will the next shot do? So sampling has an added dimension here.
So it is not quite straight forward.
We as "researchers" in shooting pellets are at the least sophisticated research level here. We are not really into the realm of "statistical significance". But, the real research question is ... does it have practical value?
If I find I can drop 5 shots under 1 inch at 50-yards 4 of 5 times using pellet "x" it has practical value to me. I have a practical experience that suggests what will happen next. It may not be statically significant but does have practical significance.
Upvote 0