The whats what on groups

[Clancy]

I am happy with any sub 1/2 MOA rifle, if you think a brand new rifle shooting 1/2 MOA (a bad group) is sub par, please show us your targets!

That’s the most recent thing I’ve worked on. Two shouldered barrels on a skinny hunting rifle. It’s a nice little setup, probably be better if I ditched the bipod for a rest and put up some wind flags. You may notice that neither of those barrels has a tendency to shoot a bad group right behind a good one, but they’ve both shot some groups I wasn’t too proud of with me behind the trigger. The barrel with the larger groups is a GT. I just got it, haven’t had time to really dial it in but it’ll get there. Those are 8 of the 12 groups I shot during my pre-200 round LD. Seating depth is at the top, rounds 3-12 from the gun. It didn’t like to be any father off than the second group. I shot the powder charges half way between those top 2. I load at the range, so I was able to go back and shoot the 38.5 load a few times with 2 different optics, and it looks like it’s going to be a winner. Really solid SD and more than enough accuracy for antelope.

 

[Clancy]

Show us 10 in a row is what hes saying

That’s all I’m saying.

Now, if you put Mr Sauter behind a good rifle on a good morning, he’s gonna make this statement from Litz look idiotic. I heard Speedy say, in an interview with Cortina, that he (himself) wasn’t a small group shooter-he just seldom shot a big one.

Myself, behind a good shooting hunting rifle when it’s 106 in West Texas, can only add validity to what Mr Litz is trying to say.

 

[CharlieNC]

… which requires a data set…

I think you (Litz) have it backwards. The group size, range, is a measure of dispersion and does not follow a normal distribution. Similarly the SD **2 (variance) follows a chi square distribution. All of these measures are bounded on the low side since it is not possible to have negative results. When attempting to apply principles of the normal distribution to a parameter which is known not to be normally distributed, it is incumbent upon the author to show a rationale regarding why this is applicable.

Over the years there have been several in depth studies aimed at "tuning" other more general distributions to best describe group sizes, something akin to custom drag curves in ballistics.

The main takeaway from the Litz paper is we all accept there is a distribution in group size when we shoot, and we must make an informed decision when interpreting the results. But there is no reason or rationale to peg SD = 30% of the average. Similarity Mr Boyer used an inadequate sample size of two shots when developing a load? Too many depends on factors!

 

[jelenko]

Hmmm. Litz works for Berger, yes?
I'm wondering if they have from their testing enough samples of shots to determine if the distribution is normal or not?

 

[CharlieNC]

Hmmm. Litz works for Berger, yes?
I'm wondering if they have from their testing enough samples of shots to determine if the distribution is normal or not?

If they have sufficient data it would be a great service to show it, as others with a bit of data have moved on from this. For example the govt utilizes mean radius, etc for their reasons. There have been a number of various approaches over the years due to lack of consensus so far. But we are known to get anal over secondary issues too!

 

[rwj]

I think you (Litz) have it backwards. The group size, range, is a measure of dispersion and does not follow a normal distribution. Similarly the SD **2 (variance) follows a chi square distribution. All of these measures are bounded on the low side since it is not possible to have negative results. When attempting to apply principles of the normal distribution to a parameter which is known not to be normally distributed, it is incumbent upon the author to show a rationale regarding why this is applicable.

Any data set can be examined by ~6 different analytical tests to determine whether it conforms to a normal distribution. If it doesn’t, so be it. If it does, then the related statistical analysis tools can be applied.

 

[Ccrider]

Show us 10 in a row is what hes saying

Exactly what I was thinking as I read this. Picking the shots that you call a group is not helpful.

 

[Doom]

A good shooter with a good rifle on a good day would be much less than 30%. Litz’s statement was a general statement covering a large sample size and Most probably a large range of shooting skills.

 

[Doom]

does not follow a normal distribution

What type of distribution do you believe would apply to the mean group size?

 

[Dave M.]

Any data set can be examined by ~6 different analytical tests to determine whether it conforms to a normal distribution. If it doesn’t, so be it. If it does, then the related statistical analysis tools can be applied.

Curious, can you explain what these 6 analytical tests are? Thanks.

 

[jelenko]

A good shooter with a good rifle on a good day would be much less than 30%. Litz’s statement was a general statement covering a large sample size and Most probably a large range of shooting skills.

Litz being a math guy, I'm pretty sure his assumption was from a solid rest - i.e., no variation in skill.

 

[CharlieNC]

What type of distribution do you believe would apply to the mean group size?

If you are a good shooter it would look something like a chi-square. Viewed as a histogram, the groups cannot get less than zero so there is a high percentage between the average and zero; with a diminishing percentage tailing to larger groups. So it is highly skewed. For a worse shooter the distribution shifts over.

Given adequate data it is not necessary to force fit a distribution, just plot a simple histogram to count percentages in different group size cells for future reference. But I'm a score shooter who doesnt find groups as interesting as individual shot statistics (mean radius and SD).

 

[jelenko]

If you are a good shooter it would look something like a chi-square. Viewed as a histogram, the groups cannot get less than zero so there is a high percentage between the average and zero; with a diminishing percentage tailing to larger groups. So it is highly skewed. For a worse shooter the distribution shifts over.

Given adequate data it is not necessary to force fit a distribution, just plot a simple histogram to count percentages in different group size cells for future reference. But I'm a score shooter who doesnt find groups as interesting as individual shot statistics (mean radius and SD).

A normal distribution doesn't have to be centered on 0, yes?

I'm thinking/assuming that Litz's statement assumed no variation in 'skill' - i.e., all shots from a solid rest.

He's making a statement about the distribution of the POI of a large number of shots.

If the mean size of the groups is, say, 0.2 MOA, 30% is .06 MOA - i.e., not very much. My understanding is there is more than that variation in group sizes shot in benchrest group shooting. While there would be skill and environmental factors involved, my only point is 30% variation in group size doesn't have to be a lot.

 

[Doom]

Just my opinion and it wasn’t a loaded question. In general I believe it would approach a normal distribution with a true half MOA rifle with a very good shooter. The more accurate the rifle the more it would approach a different distribution and Chi-Squared seems reasonable. Now when we add the average shooter variable along with environmental variables it’s anybody’s guess on a given day what it would be.

 

[Ccrider]

Variation in shooter skills can not possibly be included as part of the statistical analysis. How can you measure that?

 

[chkunz]

Something I've seen is that people take things posted above by Brian, and then they add their own assumptions to it.

It is called seeking wisdom in your own opinions.

 

[rwj]

Curious, can you explain what these 6 analytical tests are? Thanks.

Easier to let you review for yourself… https://towardsdatascience.com/6-wa...al-distribution-which-one-to-use-9dcf47d8fa93

 

[JOE360]

Lets apply Bryan's math to a slightly more precise rifle then the example. Say a gun that could average a .2 inch group, you end up with 67% of your groups being from .08in thru .32 inches and 95% of your groups being from -.04in thru .44 inches. Food for thought.

 

[BoydAllen]

Something I've seen is that people take things posted above by Brian, and then they add their own assumptions to it.

One result I have heard many say is that nodes do NOT exist and you can use any powder charge you want and with seating depth you will get it to be just as good as any other powder charge with some seating depth

It is very possible it is the case. However, if you are shooting say 22 PPC bullet from 24 lbs benchrest rifle with very heavy barrel the consistency of the load round to round is way more important than tuning to the node.

Have you heard about the secrets of the Houston warehouse?

Note in the linked article powder charge did not matter at all.

Some more readings

Secrets of the Houston Warehouse – Lessons In Extreme Rifle Accuracy

Back in 1993, Precision Shooting Magazine printed a landmark article about the findings of a group o...

precisionrifleblog.com

Rifle
View attachment 1351246

Interesting stock. Who made it?

 

[BoydAllen]

It is very possible it is the case. However, if you are shooting say 22 PPC bullet from 24 lbs benchrest rifle with very heavy barrel the consistency of the load round to round is way more important than tuning to the node.

Have you heard about the secrets of the Houston warehouse?

Note in the linked article powder charge did not matter at all.

Some more readings

Secrets of the Houston Warehouse – Lessons In Extreme Rifle Accuracy

Back in 1993, Precision Shooting Magazine printed a landmark article about the findings of a group o...

precisionrifleblog.com

You might want to watch this.