Range Report Group size means little

[Ledzep]

There's a thought I've been developing over the last several months. It goes something like this;

Spoiler alert: I've done nearly 1000 rounds worth of 30-50 shot sample size tests in .223, 6.5cm, and .300prc in the last few months and the number is somwhere between 20 and 40 depending on how much variance is in the sample.

Amazing to record POI vs. POA over 50 rounds and watch the mean point of impact move around .2-.3" at 100yd from where a 3 or 5 shot group started out.

Plotting those impacts shot by shot has me thinking that the average radius and SD of the radius from the MPOI is more telling than the extreme spread of a group of shots... and SD is really only viable if you have enough data points to show normal distribution.

 

[Steel head]

There's a thought I've been developing over the last several months. It goes something like this;

View attachment 7300721


Spoiler alert: I've done nearly 1000 rounds worth of 30-50 shot sample size tests in .223, 6.5cm, and .300prc in the last few months and the number is somwhere between 20 and 40 depending on how much variance is in the sample.

Amazing to record POI vs. POA over 50 rounds and watch the mean point of impact move around .2-.3" at 100yd from where a 3 or 5 shot group started out.

Plotting those impacts shot by shot has me thinking that the average radius and SD of the radius from the MPOI is more telling than the extreme spread of a group of shots... and SD is really only viable if you have enough data points to show normal distribution.

I agree.

My loads got a lot better when I quit chasing group size and focused on other criteria.

 

[tnichols]

I agree.

My loads got a lot better when I quit chasing group size and focused on other criteria.

Expand. I, like you, shoot a lot at distance. I’m curious what your take is. I’ll hold my reply. I try to learn something new every day,

 

[Ledzep]

Here's another way to look at it; Of all of those dots in the picture above, only 2 of them represent the extreme spread, or group size.

So we're basing the precision capacity of all of those hundreds of shots off of the two worst ones. The method gives no way to measure the "weight" of the other shots inside the outer bounds.

What are the odds that 3 or 5 random samples from all of those red dots in the picture above will contain the 2 on the outer edge? Because that's basically what we're doing when we do load development.

5x at 41.0
5x at 41.3
5x at 41.6
etc....

Then we take the best of that and load 5x at 2.850, 5x at 2.855, etc.... And then we take the best of that and call it good right?

But each time you shot 5x of a load and pass it by, you essentially took 5 of those dots above at random, and assumed it represented the population. You could have got 5 of the ones in the tightest cluster and thought it was immensely better than it really is, you could've got 5 on the outer ring and thought it was heinous, especially compared to others where you get 5 closer together.

 

[brianf]

Your venturing into statistics and curves....

Verrrrrry slippery slope, coming from a fellow math head

“He who holds the information controls the meeting”

Shave off a deviation here or there and the whole ball of wax changes.

I’m watching with very open eyes

 

[morganlamprecht]

from previous PMs, you already know my stance on this as well...tagging in to see how it goes

 

[FatBoy]

I take an entirely different approach. I take 5 different loads to 100 and shoot them into 10 shot groups. Any that are over half MOA get trashed. Any under have MOA I load 10 rounds of a couple/few different loads and go shoot them at 600 or 1000 yards. Smallest acceptable group wins, specifically looking for the smallest elevation changes, then the roundness of the group. I don't generally chase numbers if the group is inside the X ring. That said, I don't win much anymore,, so there is that.

 

[Deleted member 113831]

The ES is typically 3x the mean radius, yet 1.5x the mean radius typically contains over 2/3 of the total impacts.

I proof my chosen loads with 20 round groups at 200 yds. That has so far told the tale.

Side note: I have never managed to fire all 20 into 1 moa @ 200 yds doing this.

 

[Steel head]

Expand. I, like you, shoot a lot at distance. I’m curious what your take is. I’ll hold my reply. I try to learn something new every day,

I look for
Consistency-vertical spread
Stability-a load that won’t Bite me from heavy condensation or heat.
And finally a load that’s not picky.
I’m not sorting primers, cases and bullet or weighing charges to the nanograin.
I’m no reloading superhero but I have learned a few things in the last few years.

 

[Dthomas3523]

Here's another way to look at it; Of all of those dots in the picture above, only 2 of them represent the extreme spread, or group size.

So we're basing the precision capacity of all of those hundreds of shots off of the two worst ones. The method gives no way to measure the "weight" of the other shots inside the outer bounds.

What are the odds that 3 or 5 random samples from all of those red dots in the picture above will contain the 2 on the outer edge? Because that's basically what we're doing when we do load development.

5x at 41.0
5x at 41.3
5x at 41.6
etc....

Then we take the best of that and load 5x at 2.850, 5x at 2.855, etc.... And then we take the best of that and call it good right?

But each time you shot 5x of a load and pass it by, you essentially took 5 of those dots above at random, and assumed it represented the population. You could have got 5 of the ones in the tightest cluster and thought it was immensely better than it really is, you could've got 5 on the outer ring and thought it was heinous, especially compared to others where you get 5 closer together.

I think much of it hinges on how meticulous your loading method is.

If you’re measuring to the sorting/tipping bullets, measuring cases, measuring to the kernel, neck turning, checking all neck tension, using a force or hydro gauge, checking your BTO on most or all loaded rounds.....etc etc

Then you can be *fairly* confident your 3 or 5 shot sample is indicative of what you can expect. After all, there are very high level F class shooters doing this.

However, the average loader, IMO, is very much wasting their time with 3 or 5 round volleys for load development.

 

[Dthomas3523]

I agree.

My loads got a lot better when I quit chasing group size and focused on other criteria.

I’ve migrated over to using chrono data first, and then tuning the seating depth to that.

 

[Steel head]

I’ve migrated over to using chrono data first, and then tuning the seating depth to that.

I’m basically using the same approach now.
I look for a speed and/or a ES flat spot then see if various levels of jump helps.
The looking for ES flat spot is new to me but now that I’m looking for it it makes sense.

 

[Deleted member 113831]

I am basically looking for a load that is in the area of 2% off max. So after I find max, I try -0.2grains, 2%, +.02, +.04.

Those 4 charges will tell me what I want to know. Seating depth from there.

 

[Brandon05_88]

I agree the average radius give the better weighted measurement.

Have you seen this thread:

https://www.snipershide.com/shootin...ode-detection-technique.6996839/#post-8444830

I feel it give a proper load development technique.

 

[2aBaC̶a̶]

Eventually you'll shoot enough flyers they all group the same in the end?

Seems like what is being implied is that after 50 rds a load that puts 5 shots through the same hole at 100yds will be the same as a load that prints a 3" group at 100yds. Because group samples dont mean anything.

I find that hard to belive.

And .2 moa shift at 100 is only 2" at 1000yds. I could live with that.

 

[morganlamprecht]

Seems like what is being implied is that after 50 rds a load that puts 5 shots through the same hole at 100yds will be the same as a load that prints a 3" group at 100yds. Because groups dont mean anything.

not quite that bad...

if youre shooting the same components and getting a ragged hole group vs 3", you have other problems going on

generally when i shoot load work groups at 100 from my barrels...3 or 5 shot, the worst is 3/4-1" and the best are 1/4"-ragged holes...ive gone back out and shot both the "worst" load and "best" loads for 10-20 shot samples at the same and extended distances and they generally end up being very close to the same as the samples get larger

idk many people who shoot the "bad" loads again over multiple days...but it would likely change the way most people see things in their reloading...this is all shooting from a bipod and rear bag w/ no wind flags, although i have shot some of my rifles from really expensive rests and seen similar results

all this assumes youre only changing measurements like powder/oal/etc...changing actual components can make much larger real differences

 

[TxDerek]

I love the thought of using larger control groups at distance than the small groups at close range where shooter error could mean the difference in following a possible load or not. So many variables.

 

[Steel head]

I love the thought of using larger control groups at distance than the small groups at close range where shooter error could mean the difference in following a possible load or not. So many variables.

If your developing a load for long distances test it at long distances provided you can truthfully call your shots.
Honestly I can say I was not able to test a load at distance the first 3/4 of my long distances shooting career.

 

[2aBaC̶a̶]

Not sure an OCW would really be the same if you did it at 1000yds.

 

[Ledzep]

My experience is that a 5 shot group can print between 0.2" and 1.3" outside to outside at 100yd. A 50 shot group will print 1.0-1.3" at 100yd. Or in a better system a 5-shot will print .1-.8", and a 50-shot will print .7-.8.

I know that my 12lb rifle shoots worse 50-shot groups (.9-1.3) than my friend's BR rifle setup (.7-.95), which shoots worse groupse than a rigid accuracy fixture (.6-.7).

 

[Ledzep]

I guess my biggest point is that you shoot 5 shots of something and either keep it or pass it up based on what it tells you. My experience of continuing to shoot the same load for 50 shots has shown that doing so is in many cases entirely baseless. i.e. what the 5 shot group told you was a lie.

A 5 shot group has the ability to show you how bad a load can shoot, but is no indication (by itself) of how good a load will shoot. And even if it shows you how bad the load is, you don't know if it's representative. There's no two ways about it, even in accuracy fixtures and BR rifles the MPOI, SD, and group size (or if you take the time to measure it, average radius from MPOI) doesn't settle in until 15-30 rounds into the group.

If you seriously want to know your mean point of impact, ES, SD, dispersion potential, etc.. You need to shoot 20 rounds MINIMUM. Bare minimum. Ever wonder why a bunch of 5-shot groups are all tight, but the average location of them is different from the next group, and the next group, etc..? It's because you're randomly selecting 5 at a time of the dots in the OP picture.

ETA: Also, ever wonder why occasionally the same "happy" load occasionally throws a big group? Like normally .3-.6" but this one is .9 for some reason. And you tell yourself "Must've yanked that one" but everything seemed fine? Must've been a bad case. Must've goofed the charge. Must've seated that one different...

Must've lied to yourself and ignored data.. lol

 

[2aBaC̶a̶]

My experience is that a 5 shot group can print between 0.2" and 1.3" outside to outside at 100yd. A 50 shot group will print 1.0-1.3" at 100yd. Or in a better system a 5-shot will print .1-.8", and a 50-shot will print .7-.8.

I know that my 12lb rifle shoots worse 50-shot groups (.9-1.3) than my friend's BR rifle setup (.7-.95), which shoots worse groupse than a rigid accuracy fixture (.6-.7).

So a load that consistently shoots 1/4" groups is NOT better than a load that will consistently only do 1" groups?

Come on.

 

[lash]

So a load that consistently shoots 1/4" groups is NOT better than a load that will consistently only do 1" groups?

Come on.

That is not what he said at all.

 

[memilanuk]

There's a few good articles along these lines located here...

 

[Dthomas3523]

So a load that consistently shoots 1/4" groups is NOT better than a load that will consistently only do 1" groups?

Come on.

What he’s saying is that 1/4” gun may not be as consistently a 1/4 gun as people think.

And if you shoot four groups of 5 shots each, of you combine them all, while each one might be 1/4, in totality, they are actually .7

 

[Dthomas3523]

If you say took a shot marker system and shot 5 shots. Then replace the paper and shoot 5 more shots.

Do this for about ~30 shots, you would get what is closer to the real group size.

 

[memilanuk]

If you say took a shot marker system and shot 5 shots. Then replace the paper and shoot 5 more shots.

Do this for about ~30 shots, you would get what is closer to the real group size.

I'd rather save the string, shoot another, save it, shoot another, etc. Then export the data as a CSV to either Excel (or whatever programming environment you choose) to plot the data sequentially.

 

[2aBaC̶a̶]

That is not what he said at all.

That's what it sounds like.

Unless it's a minimum of 20 shot groups. It's not enough of a sample size to declare it valid data. So two consistent 5 shot 1/4" groups could open up to 1" and two 5 shot 1" groups could already be the extreme spread.

Basically throw the OCW test out the window unless your shooting 30rds of each charge.


I'm not disagreeing that a larger sample size isnt more more accurate, it obviously is. But a 5 should give you enough sample of a sample to at least point in the right direction.

 

[Racer88]

If you say took a shot marker system and shot 5 shots. Then replace the paper and shoot 5 more shots.

Do this for about ~30 shots, you would get what is closer to the real group size.

OnTarget TDS does this virtually. I wrote about it and demonstrated here:

https://www.snipershide.com/shootin...-calculator-composite-groups-feature.6945613/

Pretty nifty, I think!

 

[Dthomas3523]

That's what it sounds like.

Unless it's a minimum of 20 shot groups. It's not enough of a sample size to declare it valid data. So two consistent 5 shot 1/4" groups could open up to 1" and two 5 shot 1" groups could already be the extreme spread.

Basically throw the OCW test out the window unless your shooting 30rds of each charge.


I'm not disagreeing that a larger sample size isnt more more accurate, it obviously is. But a 5 should give you enough sample of a sample to at least point in the right direction.

OCW aren’t being using much anymore except for the PRS crowd. At least as far as groups for ocw.

We are behind the other disciplines quite a bit.

 

[lash]

That's what it sounds like.

Unless it's a minimum of 20 shot groups. It's not enough of a sample size to declare it valid data. So two consistent 5 shot 1/4" groups could open up to 1" and two 5 shot 1" groups could already be the extreme spread.

Basically throw the OCW test out the window unless your shooting 30rds of each charge.


I'm not disagreeing that a larger sample size isnt more more accurate, it obviously is. But a 5 should give you enough sample of a sample to at least point in the right direction.

Well, considering that the OCW method is not trying to give you the best group size anyway, you don’t necessarily have to throw anything out the window. All OCW is doing is to help you find a forgiving load that allows for many of the normal variables, be they loader or environmentally induced. In this way, it is little different than all of the other ladders or chrono tests. It gives you a starting point to begin at and then to fine tune into a good shooting load.

What Ledzep is saying is no different than what has been discussed numbers of times elsewhere. That is that five shot groups are not statistically significant if your intent is to find the best load possible.

 

[spife7980]

So how much better or worse is the worst good group vs the best worse group based on small sample when expanded to large?

 

[kthomas]

I guess my biggest point is that you shoot 5 shots of something and either keep it or pass it up based on what it tells you. My experience of continuing to shoot the same load for 50 shots has shown that doing so is in many cases entirely baseless. i.e. what the 5 shot group told you was a lie.

A 5 shot group has the ability to show you how bad a load can shoot, but is no indication (by itself) of how good a load will shoot. And even if it shows you how bad the load is, you don't know if it's representative. There's no two ways about it, even in accuracy fixtures and BR rifles the MPOI, SD, and group size (or if you take the time to measure it, average radius from MPOI) doesn't settle in until 15-30 rounds into the group.

If you seriously want to know your mean point of impact, ES, SD, dispersion potential, etc.. You need to shoot 20 rounds MINIMUM. Bare minimum. Ever wonder why a bunch of 5-shot groups are all tight, but the average location of them is different from the next group, and the next group, etc..? It's because you're randomly selecting 5 at a time of the dots in the OP picture.

ETA: Also, ever wonder why occasionally the same "happy" load occasionally throws a big group? Like normally .3-.6" but this one is .9 for some reason. And you tell yourself "Must've yanked that one" but everything seemed fine? Must've been a bad case. Must've goofed the charge. Must've seated that one different...

Must've lied to yourself and ignored data.. lol

I think in order to best test this, your rifle would have to be mounted to some sort of device that eliminated shooter influence. If you wanted to just test the load.

For those wandering groups you talk about, I'm genuinely curious how much that is caused by imperfections from the shooter, not doing exactly the same thing after every 5 shot group. Slight differences in bipod loading, cheek wield pressure, hand/finger pressure, etc. will cause a deviation in POI.

I think people are a lot less consistent at driving the rifle as they like to think they are. Small changes matter. I doubt many are coming back to the rifle EXACTLY the same way after each time they break position.

 

[2aBaC̶a̶]

Sounds crazy.

load dev could get expensive using 30 shot group samples? Not to mention barrel life.

 

[djarecke]

Sounds crazy.

load dev could get expensive using 30 shot group samples? Not to mention barrel life.

I agree. I’ve seen people shoot poorly and tell me that their load development needs work. We then spend a few minutes on breathing and follow through and all of the sudden the rifle waterlines.

These discussions are fun and academically I agree with the thoughts on small group size, sample size, etc. The weak link continues to be the shooter, and determining a load at distance with poor fundamentals just won’t work. Most well-built rifles shoot many different loads well, and about 90% of the shooters I meet could benefit from focused training on fundamentals, not a “dead nuts” load development.

 

[kthomas]

I agree. I’ve seen people shoot poorly and tell me that their load development needs work. We then spend a few minutes on breathing and follow through and all of the sudden the rifle waterlines.

These discussions are fun and academically I agree with the thoughts on small group size, sample size, etc. The weak link continues to be the shooter, and determining a load at distance with poor fundamentals just won’t work. Most well-built rifles shoot many different loads well, and about 90% of the shooters I meet could benefit from focused training on fundamentals, not a “dead nuts” load development.

Agreed.

Cool academic discussion, but at the end of the day, the shooter is almost always the weak link. You can obsess over your reloads to the nth degree, but you get to the point where it really doesn't make any difference, and spending the time perfecting your fundamentals and wind calls will go much, much farther.

Reloading is an easy thing to obsess over. But I think a more pragmatic holistic approach, is a much more effective way of increasing hits on target.

 

[308Gunner]

I have found that after load development I will usually have 1 that groups really good and 2 that are so so, I will load these same charges up and shoot them again on a second outing. Most of the time the load that was the best is still good sometimes not, I find that some days for whatever reason I don’t shoot as good as others and this just reassured me on the load.

 

[Ledzep]

I'm not disagreeing that a larger sample size isnt more more accurate, it obviously is. But a 5 should give you enough sample of a sample to at least point in the right direction.

Nope. Take a sample of 5 off of any production line and apply 6 sigma or any statistical analysis to it and you'll be swimming in scrap parts by the end of the week.

Look back at the picture in the OP. Pretend the max radius is 18", so the whole group of several hundred dots there is a 36" group at 1000yd. Barring BC variability, muzzle velocity spreads, yada yada...

That's a heinous group. 1 Mil wide.

What you're doing by firing 5 shots is RANDOMLY selecting 5 of those dots. You're not in charge. You're not picking the 5 you like, you're not picking the best 5, the worst 5... you are receiving 5 random shots as samples of the population of total shots of that load (your theoretical accurate barrel life, essentially).

So again, what are the odds that you get 5 that represent the average radius of that group? What are the odds that 2 of those 5 are the extreme limits of the total group (shots on the outer perimeter)?

What are the odds that the 5 you're randomly given are MUCH closer together than the extreme ends of the total group?

What you're doing by firing another 5 shot group is randomly selecting another 5 of that big group. So each group you fire is giving you a better and better picture of what the combo is capable of. If you don't overlay them by correlating POI vs. POA, then you lose the global reference point and subject yourself into being tricked into thinking the precision is better than it really is.

ETA: It is entirely possible that you fire a 5-shot group that is representative of the average radius of the larger group. However, of the possible groups you can shoot, THAT particular one is a lottery ticket amongst a shit load of other groups that are much better and much worse than the average.

 

[memilanuk]

Some more reading material, for those so inclined ?

 

[Ledzep]

If you really don't believe it, give it a couple boxes of ammo.

Put 20 shots into 1 group at 100yd. Any pace, best conditions you can get, just do your best to be as consistent as possible and put 20 shots of your best load(s) into a single group. Then take one of the worst performing loads you had with the same powder/bullet combo during load dev, load 20 and do it again.

My 50 person-fired (non accuracy fixture) shot groups ended up in the .95-1.5" range depending on the barrel used (factory button vs. Kreiger/Bartlein). At 20 shots they were a little better, probably had some in the .7-.8 range I can review that when I get a chance. The first 5-10 shots of all that testing varied wildly from nearly an inch down to .4". Keeping the same components/barrel they were within about a tenth of an inch after 50 shots.

 

[morganlamprecht]

I think in order to best test this, your rifle would have to be mounted to some sort of device that eliminated shooter influence. If you wanted to just test the load.

For those wandering groups you talk about, I'm genuinely curious how much that is caused by imperfections from the shooter, not doing exactly the same thing after every 5 shot group. Slight differences in bipod loading, cheek wield pressure, hand/finger pressure, etc. will cause a deviation in POI.

I think people are a lot less consistent at driving the rifle as they like to think they are. Small changes matter. I doubt many are coming back to the rifle EXACTLY the same way after each time they break position.

generally, I’d say you’re right, and I used to always like to blame myself... but a few months back I started taking multiple rifles w/ multiple barrels for each to the range and comparing things across all of them


They didn’t all show the same inconsistencies...when you REALLY get a hammer barrel, it’s hard to make it not shoot and a lot of the little inconsistencies magically go away. Ive changed bullets, powder, and seating depths randomly and they still shoot...I’ve shot them from a bipod and from a high end rest that was a buddy’s and on days with winds less than 5mph I saw no real difference at 100 yds...done powder charge work ups that span 2-3 grains and you could cover the entirety of shots from 1-25 with a penny...other barrels, they stay more in the 1/2-3/4” range, where they’ll randomly throw a 1/4” group and the next 5 are 1/2-3/4 then they may throw another 1/4” but it’s not consistent

i shot the first 3-4 groups here from the rest and the others from bipod/rear bag

same 223 barrel with a random charge of 2000mr and random jump on a 80.5..10 rounds from a barricade @ 100

this 223 barrel and 1 of my 6creeds shoot anything I put in them

Most people won’t ever really know if it’s them or the rifle until they really get a hammer one day

say all that to say this...those tiny inconsistencies in barrels make it a crap shoot getting good data from testing small samples day to day...ends up being a huge waste of time

bad part is a hammer barrel will make u hate all your others lol

 

[memilanuk]

Just for the record, I do think that we as shooters (including myself) talk ourselves into believing that load A is better than load B based on pretty thin evidence more often than not. What surprises me is how often it actually works out the way we want it to at all.

 

[memilanuk]

other barrels, they stay more in the 1/2-3/4” range, where they’ll randomly throw a 1/4” group and the next 5 are 1/2-3/4 then they may throw another 1/4” but it’s not consistent

Something that we all tend to forget / overlook is that sometimes the fliers go *into* the group, not out

 

[MadDuner]

Some days I am statistically much better than others.....
Load development never seems to matter much.
When I’m on it - Milsurp Ball ammo looks like match grade stuff.
When I’m not, the “good stuff” looks like a blind guy was shooting in gusting winds with a pistol.

 

[Ledzep]

Here's some representative MS Paint art to try to convey what I'm talking about as far as statistical viability of 5-shot groups. This assumes they all converge to the same point, which in my testing has pointed towards being the case SO FAR, but I'm not going to say that's always the case.

So essentially what you're doing with 5-shot load development is praying you have the light blue line.

 

[morganlamprecht]

Something that we all tend to forget / overlook is that sometimes the fliers go *into* the group, not out

True, for small samples that’s how you get tricked...I’m looking at multiple trips, over months and 500+ rounds in every barrel

there’s a clear difference in the better ones

 

[SmartDonkey]

I think that you are finding your limits or your rifles... not necessarily the potential of the load. I have a rifle that is very sensitive to my input, I can load the mag with 5 shots and shoot them into a 1/3" group. Reload and shoot another 1/3" inch group but it is now higher and a little to the right. Next 5 will be another good group but with a slight shift in impact. Obviously if I combine all of these groups then the overall size is going to increase-- but not because of the load. If it was the load then it was a verifiable miracle that I randomly grabbed the ones that were going to shoot together. Whatever the cause- my ability, gun heating up, mirage, scope heating etc.... it's not the load.

 

[2aBaC̶a̶]

So your basically saying any tests based off point of impact using a sample size of less than 20 are either luck or inaccurate.

Here's some representative MS Paint art to try to convey what I'm talking about as far as statistical viability of 5-shot groups. This assumes they all converge to the same point, which in my testing has pointed towards being the case SO FAR, but I'm not going to say that's always the case.

View attachment 7301483

So essentially what you're doing with 5-shot load development is praying you have the light blue line.

Not sure what I'm looking at.

Nope. Take a sample of 5 off of any production line and apply 6 sigma or any statistical analysis to it and you'll be swimming in scrap parts by the end of the week.

Look back at the picture in the OP. Pretend the max radius is 18", so the whole group of several hundred dots there is a 36" group at 1000yd. Barring BC variability, muzzle velocity spreads, yada yada...

That's a heinous group. 1 Mil wide.

What you're doing by firing 5 shots is RANDOMLY selecting 5 of those dots. You're not in charge. You're not picking the 5 you like, you're not picking the best 5, the worst 5... you are receiving 5 random shots as samples of the population of total shots of that load (your theoretical accurate barrel life, essentially).

So again, what are the odds that you get 5 that represent the average radius of that group? What are the odds that 2 of those 5 are the extreme limits of the total group (shots on the outer perimeter)?

What are the odds that the 5 you're randomly given are MUCH closer together than the extreme ends of the total group?

What you're doing by firing another 5 shot group is randomly selecting another 5 of that big group. So each group you fire is giving you a better and better picture of what the combo is capable of. If you don't overlay them by correlating POI vs. POA, then you lose the global reference point and subject yourself into being tricked into thinking the precision is better than it really is.

ETA: It is entirely possible that you fire a 5-shot group that is representative of the average radius of the larger group. However, of the possible groups you can shoot, THAT particular one is a lottery ticket amongst a shit load of other groups that are much better and much worse than the average.

I would like to see actual groups and not hypothetical simulated paint sketches.

That's just not my experience. I'm not shooting 500rds to find a a good load. I typically load 5rds of maybe 10 different charges and there is usually some clear winners to narrow it down based on groups and numbers. Those groups and numbers never vary too far from the original.

 

[Ledzep]

To that end, this thread went on a small tangent.

The point I was originally trying to convey was that extreme spread, outside-to-outside group measurements offer data only on 2 points of the 3,5,10,20, 50 you shoot. What offers better results in quantifying the dispersion is average radius (mean radius). Obviously the same statistical limitations are present in this method, which is where the 5 vs. 20 vs. 50 shot groups come in. At any rate, if you have a statistically viable sample size, and record coordinates for each shot, then get the MPOI, and calculate mean radius and standard deviation on the mean radius, you get a better picture of what you, the rifle, and the load are capable of.

So with that all said, you can set a baseline. 6.5 Creedmoor; 40.0gr RL16 with a 140 ELD-M at 2.820 fired for 50 shots produces a mean radius of .302" (group size of 1.12"). Next up we can test 41.0gr RL-16 and see that it has a mean radius of .297" (group size of 1.27"). hmm. Turns out bumping charge weight a grain didn't change accuracy very much at all.

And looking at the above, the 2nd group technically has less dispersion, but does have a larger group size. What do you put faith in? The mean radius that took 50 shots to get or the group size that capitalizes 2 of those shots? Obviously that's not always the case. Most of the time the larger group also has a larger mean radius, but they are not linearly related.

And as far as me the shooter being the weak point, each 50 shot test was made up of 5x 10-shot groups. Those 10 shot groups were often in the .5-.8" range. I have each shot logged for position in X/Y and there is no evidence of a 'wandering' zero that produces tight groups that move MPOI from group to group. There are a couple of groups here and there that seem to cluster together, but they are the exception, not the rule. I have plotted shot # vs. POI and there are no trends. There are not blocks of 10 that jump around on that 50-unit plot. The only trend I've found is velocity vs. vertical impact, and it's a pretty loose correlation, but it does seem to show up in most tests so far.

There's more that I'd like to do before I get into all of that data on here. It'll probably be another year or so before I'm comfortable making any sweeping statements about it, but for right now I am comfy saying 5-shot groups lie a LOT.

 

[Ledzep]

So your basically saying any tests based off point of impact using a sample size of less than 20 are either luck or inaccurate.

Lucky to line up with the long-term averages, yes. Inaccurate... likely. It depends on the variance of the system how bad. There are guys out there with BR rigs that probably have shit pretty tight. Nonetheless, a bunch of 5-shot groups from them will show more variance in the mean radius than 20-shot groups will.

Not sure what I'm looking at.

Several groups' mean radius as they develop. Each colored line represents a group (say each color is it's own powder charge, for example every .3gr) and as it moves to the right more shots are added to the group. The more shots you shoot, the mean radius levels out and stops moving around. The brown boxes show what the mean radius (dispersion) of the shots of each group look like at different shot counts.

I would like to see actual groups and not hypothetical simulated paint sketches.

That's just not my experience. I'm not shooting 500rds to find a a good load. I typically load 5rds of maybe 10 different charges and there is usually some clear winners to narrow it down based on groups and numbers. Those groups and numbers never vary too far from the original.

I will see about maybe formatting some of the test results to post but It's going to be a while. It's a lot of data and I'd like to A) Get more of it, and B) come up with a manner to display it so people understand wtf I'm showing.

ETA: and as far as shooting 500 rounds for load development... I don't think that's the right way to go about it. So far, and this is based off of less data than I would like to have (hence why I still want to do some more testing) I would say that a guy should pick a bullet, case, primer, and powder. Seat .025" off the lands. Pick a charge weight for the velocity he wants, then fire 20 rounds and check mean radius and MV SD. If it's under .32" mean radius and under 13fps SD, run with it. If not, change powder type or bullet. If you can't get that to work, change the barrel.