Myth Busting - 50 round shot groups at 25 yards - Extreme 9mm accuracy testing

[Dudedog]

Appropriate

1. suitable or proper in the circumstances

In a sentence.
It's always nice to have the Government or someone tell me what is appropriate since I can't seem to figure it out for myself.

Opposite
In a sentence
The mods were correct in pointing out that my reply was inappropriate.


The point I was attempting to make is what is appropriate for a scientific experiment is different than what is appropriate for an informal test.
Some tests are simple tests some are more involved.

Lets not overlook quantum mechanics.
So can any results be observed without influencing the results?
Better minds than mine have said by observing the results you have altered them.

An apology to BDS is appropriate.
BDS, sorry to have sidetracked your thread.

 

[lordpaxman]

Thanks @bds for this undertaking.

I’m hoping you prove beyond a shadow of doubt that my current reloading methods are good enough and I can stop the powder/primer/bullet/col insanity...

Your post #60 identified some variables you prioritized as controlling and it looks like a great set. Since you were considering us that use mixed headstamp brass, do you think this will be part of the testing and if so, how? When I do work ups I do use single headstamp but have no way to know if it’s the same lot, or even if that would be relevant since I’m not doing any H2O testing. I do know I went an extra step for a while and tare’d the primed case, then dropped powder, then noted the weight, and then measured the final COL. With small variations that my press gave, there wasn’t any direct correlation between differences of powder or COL with measured velocity. That lead me to conclude (ok peanut gallery, let er fly) that there was probably other variables that had a larger effect.

 

[Blue68f100]

I worked in a R&D center for 22 yrs, so I know a little about doing testing. You control as much of the variables as you can. In comes case you can not control them all, aka weather conditions. What BDS is doing and from what I have read is intent is, this is what he is going to do. Only changing 1 variable at a time. I had Million dollar budget to do some of my testing, and I would say a big part of that went toward specialized custom equipment. BDS does not have this kind of resources so he is going to make do with his best option

 

[LiveLife]

Your post #60 identified some variables you prioritized as controlling and it looks like a great set. Since you were considering us that use mixed headstamp brass, do you think this will be part of the testing and if so, how?

Yes and testing of mixed headstamp brass has already started.

When I initially started testing 9mm carbine loads, I did a mixed range brass (reloaded multiple times) vs sorted mostly once-fired .FC. headstamp brass comparison test. I expected same headstamp load to produce smaller group but the opposite happened. Comparison groups below shows mixed headstamp load with smaller core group (minus two flyers more typical of mixed range brass).

I was perplexed by the mixed vs same headstamp test so I conducted additional reloading variables tests and found the likely reason for the unexpected group results - Neck tension variance from case wall thickness (or thinness).

When I measured case wall thickness for bullet setback myth busting thread, I found .FC. and Blazer to be thinner than WIN or R.P. headstamp brass and subsequently while WIN/R.P. brass did not experience bullet setback with .354"/.355"/.356" 9mm FMJ bullets, .FC./Blazer brass experienced the most amount of bullet setback from various headstamp brass tested - https://www.thehighroad.org/index.p...nd-bullet-setback.830072/page-4#post-10926900

Based on the myth busting thread findings, I will conduct all future reference bench testing and range testing with WIN headstamp brass for consistency.

.354" - These headstamp cases did not experience bullet setback:

• CBC
• GECO
• GFL
• PMC
• PPU
• RP
• Tulammo
• WIN

.354" - These headstamp cases experienced bullet setback (After Glock 22/KKM barrel, After Glock 23/Lone Wolf barrel):

• Aguila: .000", -.001"
• BLAZER: -.001", -.001"
• BLAZER: -.001", -.003"
• BLAZER: -.002", -.046" (This is not a typo)
• .FC.: -.001", -.001"
• .FC.: -.001", -.001"
• .FC.: -.002", -.002"
• Perfecta: -.001", -.001"
• Starline: .000", -.001"

NOTE: Since CBC, GECO, GFL, PMC, PPU, RP, Tulammo and WIN headstamp cases did not experience bullet setback with .354" sized Everglades bullet, they were not tested for .3555" sized RMR 115 gr FMJ.

.3555" - These headstamp cases did not experience bullet setback:

• Perfecta
• Starline

.3555" - These headstamp cases experienced bullet setback (After Glock 22/KKM barrel, After Glock 23/Lone Wolf barrel):

• Aguila: -.001", -.001"
• BLAZER: .000", -.001"
• BLAZER: -.001", -.001"
• .FC.: .000", -.001"
• .FC.: -.0005", -.0005"

NOTE: Since CBC, GECO, GFL, PMC, PPU, RP, Tulammo and WIN headstamp cases did not experience bullet setback with .354" sized Everglades bullet; and Perfecta and Starline headstamp cases did not experience bullet setback with .3555" sized RMR bullet, they were not tested for .356" sized Zero 115 gr FMJ.

.356" - These headstamp cases did not experience bullet setback:

• Aguila
• .FC.

.356" - These headstamp cases experienced bullet setback (After Glock 22/KKM barrel, After Glock 23/Lone Wolf barrel):

• BLAZER: .000", -.0005"
• BLAZER: .000", -.001"

When I do work ups I do use single headstamp but have no way to know if it’s the same lot, or even if that would be relevant ... there was probably other variables that had a larger effect.

What I found that was more significant (Had greater reloading variable affect to overshadow other reloading variables) was resulting bullet setback (or no bullet setback) from case wall thickness/brass quality.

So simply using same headstamp brass won't ensure your groups will get smaller. You need to use same headstamp brass that won't produce bullet setback with particular bullet you are using.

 

[LiveLife]

I worked in a R&D center for 22 yrs, so I know a little about doing testing.

You control as much of the variables as you can. In comes case you can not control them all, aka weather conditions. What BDS is doing and from what I have read is intent is, this is what he is going to do. Only changing 1 variable at a time.

Yes, that is what I will attempt to do with the prototype machine rest IF it produces more consistent groups at 25/50 yards.

I had Million dollar budget to do some of my testing, and I would say a big part of that went toward specialized custom equipment. BDS does not have this kind of resources so he is going to make do with his best option

I sure don't. It would be nice to have a dedicated 100 yard indoor range with concrete floor that I can anchor a nicer hard machine rest but I don't. Instead, I gotta build a prototype "portable" (Yes, portable) machine rest ... wish me well.

And without further ado, here's the start of my prototype machine rest build thread - https://www.thehighroad.org/index.p...-pcc-machine-rest-build.852174/#post-11147471

 

[GE-Mini-Gun]

The thing with a Ransom is it shows the potential of the gun and ammo...not the shooter. I've always wondered why people say...I get .5 MOA 3 shot groups out of X rifle...3 shots does not a group make

 

[Typetwelve]

The thing with a Ransom is it shows the potential of the gun and ammo...not the shooter. I've always wondered why people say...I get .5 MOA 3 shot groups out of X rifle...3 shots does not a group make

Yeah, pretty much. I'll admit that I'm one to bite on things like online tests, but I'm slowly learning not to. I think a mass majority of people's (or test's) claims of accuracy are not true. I've come to believe that if I can keep my short range pistol groups to less than 3", my 25yd to less than 6", and my 50yd to less than a paper plate, I'm more than pleased. If someone can do better, and I'm sure they can, good for them.

 

[LiveLife]

I think a mass majority of people's (or test's) claims of accuracy are not true.

Yes I agree and reason for this thread.

IMHO, I believe 3 shot groups are subset of 5 shot groups, 5 shot groups are subset of 10 shot groups, 10 shot groups are subset of 20 shot groups, etc.

For us to validate true accuracy of a load with a particular pistol, large enough sample size is needed that does not increase in size past a certain number of rounds. If the group size continues to increase with increasing number of rounds, it's still a subset and not a true reflection of accuracy.

And that's what I am hoping to accomplish with this thread - Keep shooting with machine rest until group size does not increase.

 

[fxvr5]

For us to validate true accuracy of a load with a particular pistol, large enough sample size is needed that does not increase in size past a certain number of rounds. If the group size continues to increase with increasing number of rounds, it's still a subset and not a true reflection of accuracy.

And that's what I am hoping to accomplish with this thread - Keep shooting with machine rest until group size does not increase.

While this might seem possible, it's not a realistic possibility. In practical terms, the round count will end up in the hundreds or thousands. There is some truth to the notion that more shots increases group size.

 

[LiveLife]

round count will end up in the hundreds or thousands.

Not practical for me.

There is some truth to the notion that more shots increases group size.

That's what I found with some loads from 3 vs 5 vs 10.

 

[fxvr5]

That's what I found with some loads from 3 vs 5 vs 10.

And it's the case for 25, 50 and 100 shot groups, too.

 

[LiveLife]

And it's the case for 25, 50 and 100 shot groups, too.

 

[rpenmanparker]

OP, I disagree with your reasoning on larger groups being valuable when using group size to judge precision. You are making a fundamental statistical error. When a rare flyer is on the paper it produces the same size group as if many of the shots fly that wide. You are making sure to capture flyers in your groups when they don’t contribute that significantly to the results in a statistical sense. The more statistically sound method is to shoot your 50 shots as 10-5 shot groups and average the 10 groups size results. That captures the effect of the flyers, but doesn’t overweight them. Especially if you also calculate the standard deviation. The youbwill also get a sense of the contribution provided by the flyers.

If you want to shoot all 50 shots as one group, then you shouldn’t judge precision by group size, but rather by scoring and adding up each shot to get a total. Then divide by 50 to get the average score per shot. Problem is with 50 shots you will surely get many of them going through holes made by earlier shots. You won’t be able to score those very well. They will he lost. I like the average group size method better.

 

[fxvr5]

But the average group size can be misleading because it can misrepresent the aggregate (50-shot) group size. The 50-shot group illustrates the dispersion of shots. Separate subsets (like ten 5-shot groups) might not overlap very well and can give a misleading representation of the dispersion of shots.

It is possible to have a smaller average group size with a load, even when its aggregate group is wider than other loads. Hence the problem with averages.

The only flier that can be excluded from the measurement are "called" fliers. That's why a mechanical machine rest is the preferred method. It has no called fliers. Every shot counts*.

* The exception is first round fliers from a semi-auto pistol where a manually cycled slide might not put the barrel in the same position as a fired cycle. But a wise tester shoots the hand-cycled round off target.

 

[rpenmanparker]

But the average group size can be misleading because it can misrepresent the aggregate (50-shot) group size. The 50-shot group illustrates the dispersion of shots. Separate subsets (like ten 5-shot groups) might not overlap very well and can give a misleading representation of the dispersion of shots.

It is possible to have a smaller average group size with a load, even when its aggregate group is wider than other loads. Hence the problem with averages.

The only flier that can be excluded from the measurement are "called" fliers. That's why a mechanical machine rest is the preferred method. It has no called fliers. Every shot counts*.

* The exception is first round fliers from a semi-auto pistol where a manually cycled slide might not put the barrel in the same position as a fired cycle. But a wise tester shoots the hand-cycled round off target.

Here is why I think you are wrong. You could have 50 shots defining a circle 3 inches around. All the shots around the circumference. Alternatively you could have 49 shots through one hole and one shot 3 inches away. The group sizes would be the same, but the precision of the cartridges would be very different. You get much better information from averaging smaller groups or scoring each shot with the values from the target rings in a much larger group and averaging them. In the example I just gave that one flier would average down to give a very different result from the 50 fliers. You surely can see this. I did this stuff for a living for 40 years. I know what I am talking about.

One thing more needs to be said. The whole idea of using statistically sound methods is to get the most information with the least effort. Your idea of continually adding more shots to your group, besides giving s wrong result, flies in the face of this scientifically proven concept.

 

[rpenmanparker]

Yes I agree and reason for this thread.

IMHO, I believe 3 shot groups are subset of 5 shot groups, 5 shot groups are subset of 10 shot groups, 10 shot groups are subset of 20 shot groups, etc.

For us to validate true accuracy of a load with a particular pistol, large enough sample size is needed that does not increase in size past a certain number of rounds. If the group size continues to increase with increasing number of rounds, it's still a subset and not a true reflection of accuracy.

And that's what I am hoping to accomplish with this thread - Keep shooting with machine rest until group size does not increase.

Your belief about this is statistically false. Someone as interested in this kind of thing as you are really should learn something about statistics.

 

[fxvr5]

Here is why I think you are wrong. You could have 50 shots defining a circle 3 inches around. All the shots around the circumference. Alternatively you could have 49 shots through one hole and one shot 3 inches away. The group sizes would be the same, but the precision of the cartridges would be very different. You get much better information from averaging smaller groups or scoring each shot with the values from the target rings in a much larger group and averaging them. In the example I just gave that one flier would average down to give a very different result from the 50 fliers. You surely can see this. I did this stuff for a living for 40 years. I know what I am talking about.

But we're not interested in the average. A test of how a gun shoots the ammo is the full dispersion of shots. The 50-shot group tells you that.

 

[rpenmanparker]

But we're not interested in the average. A test of how a gun shoots the ammo is the full dispersion of shots. The 50-shot group tells you that.

That is what standard deviation is for. You are hopelessly out of your depth.

 

[fxvr5]

That is what standard deviation is for. You are hopelessly out of your depth.

I have data (revolver in a Ransom Rest at 25 yards) that contradicts your premise.

My data shows a load (load A) with a larger group average and larger standard deviation than load B. With your approach of using averages, load A is less accurate than load B. However, load A has a smaller aggregate group than load B.

You're out of touch with the real world.

Averages can be misleading. Total dispersion tells the truth.

 

[rpenmanparker]

That is what standard deviation is for. You are hopelessly out of your depth.

I shouldn’t have said that. I apologize. But you really have no idea how to use data to know what a good cartridge load is. It is like you are making US autos in the 1970s and trying to compete with Japan. You just don’t know how to judge a good load. You can’t just keep shooting more cartridges hoping the next one won’t be a flier. The fliers will occur with a frequency consistent with their probability. You have to find out what that probability of a flier is from the average and standard deviation. Not from continuing to add shots. You can’t shoot an infinite number of cartridges. But you can learn how to predict the frequency with which any particular shot position will occur.

 

[LiveLife]

OP, I disagree with your reasoning on larger groups being valuable when using group size to judge precision. You are making a fundamental statistical error. When a rare flyer is on the paper it produces the same size group as if many of the shots fly that wide. You are making sure to capture flyers in your groups when they don’t contribute that significantly to the results in a statistical sense. The more statistically sound method is to shoot your 50 shots as 10-5 shot groups and average the 10 groups size results.

When I started 2019, I told Walkalong I hoped to post more factually and objectively and then reviewed what I had done for reloading consistency as I approached range testing phase of many myth busting threads of past year. I feel good about many myth busting threads that either confirmed or busted previous notions about reloading variables and consistency - https://www.thehighroad.org/index.p...-and-discussions.778197/page-10#post-10966692

Those that have followed my load development threads have seen my powder work up target groups and KNOW I post the ACTUAL targets and holes they produced. As Walkalong has expressed in the past, data collection can be tweaked or continued until desired results are achieved. I don't tweak my myth busting thread data or range test data. I post whatever measurements I collect and allow holes on target to speak for themselves. Such collection of data is this on case wall thickness and subsequent bullet setback using various diameter bullets (.354", .355", .3555", .356"). Factual numbers don't lie - https://www.thehighroad.org/index.p...neck-tension-and-bullet-setback.830072/page-4

But as I approached building of prototype machine rest and eventual range testing of various myth busting thread findings, I held an unsettling feeling about range test results of past years where 5 and 10 shot groups were used. What prompted me to start this thread was because I HAD ALREADY DONE 20-30 round group testing with scoped pistol caliber carbines using bipods/rifle rests (to lessen human factor) and various 22LR ammunition with my 10/22 Take Down and ARs with CMMG conversion kits. As round count increased from 5, 10, 15 and 20+, I saw some groups remain relatively small but some groups started to increase in size. Well, I immediately thought of "shooting variable" of barrel heating up and become more flexible. BUT why would that occur for some loads and not others, especially with a HOT BARREL?

That's when I pondered the possibility of smaller round count group size being a subset of larger round count group with INITIAL flyers being part of that larger group.

So how do we rule that out?

Shoot more groups. And I did.

When I got similar results, I decided to start this thread.

So if you disagree with my reasoning of larger groups being valuable, that's my reasoning. If you want to talk about fundamental statistical error, keep in mind my previous range testing were done with mixed unsorted range brass with unknown reload history or condition of brass. Those reloading variables will muddy up the statistical error factor quite a bit. And hence why I conducted so many myth busting threads to make reloading variables more consistent as outlined in this post - https://www.thehighroad.org/index.p...-and-discussions.778197/page-10#post-10966692

If you want to shoot all 50 shots as one group, then you shouldn’t judge precision by group size, but rather by scoring and adding up each shot to get a total. Then divide by 50 to get the average score per shot. Problem is with 50 shots you will surely get many of them going through holes made by earlier shots. You won’t be able to score those very well.

When the proof of concept is successful with the prototype machine rest, I can do cumulative vs separate 10 shot groups on grid as I already posted previously.

Your belief about this is statistically false. Someone as interested in this kind of thing as you are really should learn something about statistics.

I am not a statistician. I am more "Holes on target speak volume" kind of a guy.