Information content in a single 3-round group?

[atek3]

Most handloaders try to pick the "most accurate load" for their rifle.

Generally they vary the powder weight and assemble "testing ammo."

I got a batch of test ammo from a custom handloader. The instructions read:
"fire a 3 round group of each of the 4 test batches. Pick the most accurate load. If two tie, fire the remaining two rounds of each batch to break the tie.

Can anyone guess the likelihood that this procedure will choose the most accurate load?

Isn't it most likely to pick the load that happened to have the most flattering 3-shot group?

Related question.

Suppose I gave you load A and load B. How many rounds of each would you need to be 95% confident that one load was more accurate than the other?

Certainly there are some ballistic statisticians among highroaders, right?

 

[notaglockfanboy]

I usually find the best load/loads then I load them in batches of 5 and reshoot. That has worked for me and by shooting 5 instead of 3, that gives me a little wiggle room for fliers caused by me. If I am getting four consistently together I will stick with that particular load.

 

[Hardtarget]

I have been playing with reloads for my AR. I run several loads and head to the range. What I do is shoot three round "mini groups" to see if one group stands out with potential. Then I play with five round groups above and below that loading.

However, I think I'm just kidding myself a little bit in that I'm not a competition shooter or a long distance varmit/target shooter. If it looks good at 100 yds I'll shoot it out to 200 yds. I need to get to a range with a bit more distance to see if my reloads really are any good. Its just couriosity because my club only has 200 yds max.

I just use the 3 shot group as a starting point. It needs more work to prove itself.

Mark

 

[oldreloader]

I'm in the process of changing from AA2230 ( I can't find any AA2230) to BL-C(2). Just loaded up 3 each at 26.0,26.5, and 27.0 gr. Whatever groups best I will load 5 each of it and .2 grains on both sides and go from there.

 

[gamestalker]

I also recommend shooting off a good quality rest, lead sleds are pretty decent. This helps to reduce shooter induced fliers, thus delivering reliable trends / evaluations.

Something I may not have understood, are you asking about someone else's development for their rifle, to be fired from your rifle? Or did I miss something?

GS

 

[rskent]

Sorry, I just don't get the three round group idea. If your gun shoots one inch and your hold is one inch, your three shot group is as likely to be one hole as it is a two inch group.

I think if you want to find out how your loads shoot, you need to shoot some loads.

Sorry, I just don't get it.

 

[Walkalong]

The instructions read:
"fire a 3 round group of each of the 4 test batches. Pick the most accurate load. If two tie, fire the remaining two rounds of each batch to break the tie.

Can anyone guess the likelihood that this procedure will choose the most accurate load?

This is assuming the shooter is a machine and will be perfect every time. It doesn't happen that way for 99% of us.

Shoot groups with each charge weight. Do it again. See if there is a pattern. Do it again and if the pattern persists start honing it in.

Those at the top of their game who are shooting all the time can do it much faster. Take some time off, not so much.

 

[Dave P]

If you are a very good shooter, and you want confidence your load is good, shoot a Minimum 5 rounds, and 10 is much better.

 

[rsrocket1]

I don't think 3 round groups will tell you that a load is good, but it will might tell you that a load is not worth keeping. I always start with 5 rounds of each load. If they are all over the place, it shows that they are bad. If they group well, it warrants further testing.

 

[witchhunter]

I load 9 of each load and shoot 3 groups of 3 shots each and average them. I usually go .5 grain increments. When I pick one, then I will try .1 for a few on each side of the node to fine tune. Then work with different primers and seating depth for the final recipe.

 

[mdi]

I'm not the greatest shot and 3 rounds just isn't enough info for me. I really like reloading so if I make "test batches" of 8 or 10, so much the better. I once got a 5 shot 1 1/4" group with my Garand at 60 yards (I know that's short but it's all the room I had). Now was that load accurate? I really doubted it, so I ran 16 (two clips) more to check. Not surprisingly, the groups opened up to 3"+. For my guns and test loads I make up 12 rounds for my revolvers (two cylinders full), 8 rounds for my American (two magazines full), 16 rounds for my Garand, and my 1911s get 14 rounds. I get a much better average that way...

 

[Grumulkin]

I am a good shot from the bench and am very consistent. In regards to shooting, I don't claim that a gun will do such and such "if I do my part" because I consistently do my part and shoot better from the bench than a lot of guns do.

The above 0.042 inch three shot 100 yards group looks pretty good doesn't it? The trouble is that it was done with a barrel that could not be counted on to do better than 1.5 to 2.5 inches consistently. That group was, you might say, fantastic luck.

So, while I do use three shot groups for load workup with it's significant savings in costs of reloading components and barrel wear, I don't feel a single three shot groups proves a load is good. If I get a good three shot group I'll then do a five or ten round test or perhaps several three shot groups.

 

[30Cal]

It depends on the rifle. I have a couple rifles that won't shoot well much past 3 shots. So I try for at least 2 three shot groups. If the rifle will go 5rds, then I'll shoot 5rd groups.

For my heavy match rifle, I go with a 10 shot group. It's immediately clear which is best at that point.


But if you're only choosing between 4 loads, the 3 shot group will probably put you on the best one most of the time.

 

[Jim Watson]

I am not a statistician, but if you can find the old NRA books on handloading, there are articles on group size and number of groups to accurately measure performance.

General Hatcher wrote that Springfield Armory machine rest tested National Match 1903s and Springfield Sporters, and sent the test target with the rifle. Naturally you would be very proud to get a tiny "bumblebee" group with your rifle and would show it off at any opportunity. Hatcher realized that these tiny groups were statistical anomalies and would have the rifle reshot. The group would always be near the average for the ammo. The owner might not be as excited, but he would have a better idea what his rifle would do in the long run.

 

[Sunray]

Suggest you re-think "from a custom hand loader". Kind of defeats the purpose of reloading. Going to take eons to get your load sorted out.
However, 3 rounds is an old standard. A mag load(not 20, if that applies.) or 5 is better. Group size isn't as important as getting consistent groups. Hard to time wise, if you aren't doing the loading.
There should be no 'ties' either. 1/10th of a grain is enough to go from an inch to one hole. If your "custom hand loader' has done his job properly, each load will be slightly different.

 

[Jlr2267]

From a statistical POV, the answer is that there is no single apriori answer to the question of how many samples to achieve x% confidence. It depends on how much variation you see from each load...

You can use the F-test to test for unequal variances.

 

[Grumulkin]

1/10th of a grain is enough to go from an inch to one hole.

I think that the only way that could be true would be in a case with a pretty small capacity or a very touchy powder which I wouldn't be excited about using.

 

[1066]

An interesting and amusing analysis of the value of three shot groups here:

http://www.ar15.com/mobile/topic.html?b=3&f=118&t=279218

 

[BSA1]

The problem I have is I never know if I am testing the ammunition, the gun or myself. So the only solution is to repeat the testing over a period of several range sessions on different days.

For example I know from repeated range sessions that my Ruger .270 will consistently shoot 1/2" 3 shot groups at 100 yards with it's pet load. From this I know that the factor not doing it's job when it doesn't group well is the one pulling the trigger.

 

[rduckwor]

Three round groups represent luck, five rounds groups represent a good idea, ten round groups represent the truth. About the load, the gun, and the shooter.

RMD

 

[bush3755]

One 3-shot group shows where each one went. The odds of a second one being the same size and its center where the first one was is slimmer than a sheet of letter paper viewed edgewise.

Take a 50-round box of 22 long rifle, shoot two shots to get a zero at 100 yards then shoot sixteen 3-shot groups atop each other with the same sight setting on one target. See if all are the same size as the first one and all centered at the same place. If they are, then 3-shot groups are good for you to use. Pay close attention to that 48-shot group.

 

[flat six]

My statistics classes are a distant memory and I may be applying the logic incorrectly to this situation.

However, for what it is worth, I believe that as a very rough approximation, to have 95% confidence that a 1" group that you shot is truly between 0.9" and 1.1" (in other words, 10%, plus or minus) would require almost 100 rounds. If you were willing to accept that the 1" group you shot was truly between 0.8" and 1.2" (20% +/-) and not just a bad sample, then you would need about 20. 11 rounds would give you 95% confidence of +/- 30%.

Using this same logic, 5 rounds gives you about 45% confidence interval and 3 rounds gives you about a 55% confidence interval (whatever group you actually got, plus or minus 55%) at a 95% confidence level.

This is all based on percentages, so you could take a 2" group and at 10% confidence interval (about 100 rounds) and assume it is between 1.8" and 2.2". If you shot a 2" group with three rounds, you could be 95% sure that it is actually between 0.9" and 3.1" (2" +/- 55%). If you got the same 2" group with 5 shots, you could be 95% sure that it is actually between 1.1" and 2.9". An 11 round group would be 95% likely to fall between 1.4" and 2.6" (+/- 30%). All this is a rough approximation with assumptions and rounding....

Basically, you can never be 100% sure, unless you fire all of your ammo. You just have to decide how much uncertainty you can live with.

Hope this helps. If anyone can correct my logic, I'd appreciate it.

 

[TimSr]

Can anyone guess the likelihood that this procedure will choose the most accurate load?

Isn't it most likely to pick the load that happened to have the most flattering 3-shot group?

You tightest group is your most accurrate load, but the same test needs to be repeated for a trend.

It's a good start. It's only a start. Depending on how different the loads are, the idea of the test may just be to eliminate those combinations that don't work, and improve on the best one/ones after this test.

 

[Nature Boy]

What would be the difference in data gathered between 10, 3 shot groups or 3, 10 shot groups.

I usually do 4, 5 shot groups, but sometimes do 10 shot groups. at the end of the day, the data has to speak to you.

 

[bush3755]

Nature Boy,

Shoot a 30 shot group then pick three different 10 shot hole groups. Or pick ten different 3 shot hole groups. See what their sizes are as well as where their centers are relative to each other.

Doesn't matter if the shot holes are picked in the order they were fired. If shots start walking due to barrel heat, you'll need to let it cool down to ambient temperature every few shots. Patience is a virtue in this test.

One 20 shot group is better than any combination of 4, 5 or 10 shot ones.

Arsenals test 30 caliber match ammo shooting over 200 shots into a single group at 600 yards. Some 3 to 5 shot strings have a spread of a couple inches. The whole group is over 12 inches.

Bench rest 5 shot group records are a small fraction of aggregate records that are the average of several groups. The largest group shot in aggregates is bigger than the average.