Are We Doing Load Development Wrong?

[Beiruty]

In the benchrest shooting,
If the shooter is chasing extreme accuracy.

He will drop powder to one kernel
He will measure BTOG of the round to within 0.002"
He will measure BTOG of the bullet to within 0.002"
He might weigh the case volume too and bin to 0.5gr or less
He would measure TRO to within 0.002"

What is he tuning for? What variables left to tune for?
1) Weather changes
2) Powder burning consistency?
3) Primer burning consistency?
4) "name your variable"

 

[Dave M.]

The reality of it is, in order to shoot what Bryan and others might deem statistically significant number of shots per group, we’d likely shoot our barrels out before we ever made it to the first relay of the first competition we took the barrel to. Also, can you imagine the cost to do so????
Dave

 

[Coyotefurharvester]

An ammo manufacturer is going to say charge weights and seating depth don't matter, because 1. they only use one powder charge weight in a production run 2. Their charge weights are not as accurate(consistent) as a hand loader can be. Seating depth is determined(for ammo manufacturer) by 1. SAAMI specs, not by the quality of group's on a target. 2. Because seating depth consistency is not as accurate with new brass, due to the friction between bullet jackets and case mouths.

 

[chkunz]

This is a copy/paste from a previous post.

In load development we frequently want to know if load A is more accurate than load B. Statistical analysis can help provide an answer to this question. The number of shots required for a meaningful answer depends on the statistical variance between A and B and the statistical confidence level that you choose. The smaller the variance between A and B the more data that is required for a meaningful answer. Also, the higher the statistical confidence level you desire the more data that is required. A statistical confidence level of 95 means that the probability is 95% that A is different than B but that also means that there is a 5% chance of being wrong. While a 98% confidence level has only a 2% chance of being wrong. The principles of statistics will guide you as to how much data you need to achieve your choice in confidence levels.

Although a basic understanding of statistics and statistical analysis is helpful in group analysis, books on the subject are not easy to master. If you have a good high school or college background in math, give the subject a try or better yet find a friend that can teach you.

I have done considerable load development for National Match Course competition (aka XTC). For load development for this application, I shoot 10 shot groups. I see where others doing load development for various applications report mostly using three, five or sometimes seven shot groups but rarely 10 shot groups. Their approach to load development may be appropriate for their application however using insufficient data for analysis for any application can lead to incorrect conclusions.

It has been shown that for long run averages 20 shot groups can be expected to average 1.79 times as large as three shot groups and 1.45 times as large as five shot groups but only 1.17 times as large as 10 shot groups, therefore, I choose 10 shot test groups for XTC load development.

While shooting XTC load development groups, I have observed the hits as the groups develop and sometimes the first three shots are very close together but after all 10 shots are fired the group is usually just another average 10 shot group for that load. Sometimes this is true for the first five shots as well but not as often as for three shots as would be expected. I think this supports my decision to go with the 10 shot test groups rather than three or five shot test groups.

There is also a tendency to throw out “flyers” when firing groups for accuracy testing. Sometimes I see wide shots that at first look like flyers but after the 10 shot group is in place many times the flyer no longer looks as much out of place. So, I do not ever throw out any flyers. Once I think I have a pet load defined, I shoot multiple ten shot groups with that load and average them and then if any wide shots are indeed not representative of the load, they do not play a significant role in analysis and decisions.

I calculate group size using the “extreme spread” method and also the “mean radius” method. The mean radius calculation values all shots in the test group on an equal basis as opposed to extreme spread which only values the two extreme shots. Extreme spread is more widely used for group analysis, but mean radius is a superior measure of merit and has significant advantages such as reducing the significance of a potential flyer.

 

[Bill Norris]

It has been said that 3 or 5 shot groups test the rifle & load and 10 shot groups test the shooter. I totally agree with that statement. But, once the load has been chosen I like to try that 10 shot group at some point just to see if I can still hold 1/4 inch because if the shooter cannot hold 1/4 inch then he/she cannot shoot 1/4 inch.

 

[Beiruty]

In the benchrest shooting,
If the shooter is chasing extreme accuracy.

He will drop powder to one kernel
He will measure BTOG of the round to within 0.002"
He will measure BTOG of the bullet to within 0.002"
He might weigh the case volume too and bin to 0.5gr or less
He would measure TRO to within 0.002"

What is he tuning for? What variables left to tune for?
1) Weather changes
2) Powder burning consistency?
3) Primer burning consistency?
4) "name your variable"

The matter of fact,
reading the wind and shooter's shooting skills are way more important than all the above.
Have you all considered
Recoil energy as a merit factor for accuracy?
Muzzle energy vs total rig weight for accuracy?
The use of benchrest sled vs bipod, or just the use of a backpack?

 

[CharlieNC]

How much does 0.1gr charge affect velocity? Not much you say, and the SD is large enough that it will require 20 shots each in order to have statistical significance in order to know how much the 0.1 matters, and we can do a t-test to assess the level of significance! Or you can load two each at +/- 0.2 and 0.5 and quantify the effect even better and learn more about what's going on as well. Substitute group size for SD, same principle. Target results also respond as a smooth trend, there is not a unique "outlier" that is magically better vs the overall behavior.

 

[cdgaydos]

The flaw in this, for us, is 'Variation is everywhere'.
You can't calibrate with every adjustment at the same time. Instead you use a calibration procedure that includes isolating prerequisites, and then takes you through coarse adjustments 1st, then fine. If a long analog string, then grooming(repeating) may be in order, but it's still in logical order.
Now you might have arrived at your procedure with DOE. That's good stuff, but a different phase.

It was mentioned that Bryan considers 1/2gr under 'max' as assumed developed.
Hard to believe, but it might explain inconsistent results across 20sht strings..
Max itself is a local matter to be tested for, while having different meaning from different builds.

My point in “variation is everywhere” is that in any process, changes in inputs have and effect on the output or outputs. Changes can come from purposeful manipulation or random noise (both).

We cannot control everything and our ability to control is limited by capabilities and even our abilities to measure… it’s never “perfect”.

Understanding the components of variation and the magnitude of the effects of that variation is critical.

Further, while statistics are a powerful tool, in any process improvement effort one must distinguish between statistical AND practical significance.

A statistical approach is not by definition an approach requiring significantly more “shots”… statistics are not in conflict with lower shot count approaches… statistics simply aid in determining the power of a test and help with estimating certainty and assigning “risk” (being wrong).

Variation is everywhere… but, it is also relative. For some, they and their equipment have reduced variation to extremely low levels (relative to practical requirements) and they can “see” the effects of load development changes with few shots. (Power and Sample Size)… and most shooters who compete have a lot of previous data that helps put things “in perspective”… they know what good looks like…

The statement “variation is everywhere” does not imply nothing can be done and give up… or nothing can be inferred without a tremendous amount data. It just means that one should be aware of and consider the sources and magnitudes of variation before making inferences… of practical significance.

 

[STS]

They wrong.

LOL This whole thread could have stopped after Alex's response.
Edit to add: And he did it without the apostrophe and the 're'. Saved three key strokes.

 

[Beiruty]

In science, if you do not postulate a theorem and prove it, you are stating an opinion.
If you would like to "observe" or "measure" an outcome of an experiment, you need to rely on statistical analysis or your "observation" or "measure" would be shown to be unreliable and inconclusive.

 

[Idaho-Lefty]

I think people that sell bullets would love you to shoot 20 shot strings for load development.

In other words. Do what works for you since there will likely be no widespread agreement on this one.

THIS says IT,. all ^^^
So many types of Shooting ( Benchrest, PRS, Hunting Big Game L R and Varmints to, name a Few ) and, what's "acceptable" to One guy may NOT, be for,.. the Next

 

[G Liles]

I think people that sell bullets would love you to shoot 20 shot strings for load development.

In other words. Do what works for you since there will likely be no widespread agreement on this one.

THIS ^^^^

 

[gunsandgunsmithing]

You don't have to fire that many rounds if the results are already predictable. IOW, if I make a change and it does exactly what it should do, that covers a ton of ground. My tuner test is based on just that. I look for predictable, repeatable results on both sides of a sweet spot. In reality, rather than looking at individual groups, I look at, at least, the two groups before and after the smallest group, along with poi change. So I'm actually looking at five groups, not one, as well as poi. I'm not gonna try to calculate the statistical significance of five groups that do exactly what each SHOULD DO, but I'd say it's exponentially better than any single group, even a 20-25 shot group.

 

[cdgaydos]

I’d like to see them come to a short or long range BENCHREST match. Especially a short range match where you need to shoot a different load nearly every time you sit down out the bench. I think it would change their reality.

I’d offer that you are able to make those changes every time you sit down because you have compiled a tremendous amount of data… you know what to change based on conditions. Your “load development” has been ongoing… I’d suggest that at some level of perfection (short range benchrest) you can’t do load development with any number of shots and load for a season… because the source of variation is uncontrollable on any given day… but, the level of precision in the shooting system allows for meaningful inferences from fewer shots. The statistical significance comes from the precision of the tool and shooter as well as from ongoing development on match day.

For service rifle (as an example where there is notably less precision) development with 3 shot groups is risky.

(There is no universal answer to this question… but, I feel compelled to defend statistics…)

 

[Dusty Stevens]

Yall, please do this if youre an actual competition shooter. Follow their words exactly.

 

[1kbrxDave]

At this year’s IBS short range Group Nationals, I changed my load almost every other target, and by doing so I won the 3-Gun, 2-Gun, 2 Grand Aggs, and 3 yardage aggregates. In all due respect for the experts, to be on the ultimate edge of accuracy and precision you have to make load and maybe bullet seating depth changes. Additionally, I was blessed with good luck and the spirits of those that have passed before me!
Happy Holidays to all of you.
Lee
View attachment 1394726

Well to me a Picture is worth a 1000 words. very nice sir

 

[jelenko]

In your case it may be true, BUT just because you shoot that much to confirm an accurate load does not mean everyone else should and would have to in order to find their accurate load. No two rifles and loads are alike.

Not sure exactly what chkunz is meaning, but I do a similar type of load development. With 2 MOA 10 rings and 1 MOA X rings, I do enough load development to get a 5 shot group that's close to 1/2MOA. The follow on 10 shot groups are shot in competition. I use these results to confirm/assess the load I've chosen.

 

[drop_point]

One thing is for sure, I can shoot a 3-5 shot group, and it won't get any smaller...

Load development is sometimes about eliminating what doesn't work as much as it is identifying what does work.

When I am developing loads, I try it three or four times before "settling" on it...and its always a candidate for modification later on.

 

[cdgaydos]

It’s all about noise and confidence. If you want to be say, 95% confident that the population mean (say 5k rounds - life of a cut rifling barrel give is take) of the charge you settle on will be representative within X margin of error for a given standard deviation then you can do some math and calculate the sample size needed. This of course assumes continuous data that follows a normal distribution. Which our stuff doesn’t. The shooting platform (shooter) varies, barrel changes over time, as does the atmosphere and probably powder lot and bullet geometry unless you buy a huge amount of a single lot…. That is why from one outing to the next your realistically have days of big groups and days of small groups.

Find a load. Go shoot it a lot and then do the math on all of those trials with whatever sample size of each and you will end up with a population mean and standard deviation accounting for all of the variation.

I have a table somewhere I generated that tells me sample size and confidence intervals for desired margins of error and sample standard deviation. I just filed it away because it was saddening and I will never fire the # of rounds needed to be happy.

To be honest, I fire at a big target and I hold my rifle and use a sling… I am by far more error input than internal and external ballistics of my rifle. So finding a node where elevation is stable and j am not redlining and then adjusting to find smallest group (and I am not using .003 increments) and reverifying with larger sample size and ultimately at match in 20 shot strings is my general method.

You want to know what the rifle and load will do then go find a setup like the smallbore ammo test facilities… otherwise spray and pray.

Yep. Statistical vs practical significance… and know when it’s “good enough”.
I’m new to service rifle and am realizing that my previous load development methods may be getting in the way of improving my skills…

 

[DR_2297]

Yep. Statistical vs practical significance… and know when it’s “good enough”.
I’m new to service rifle and am realizing that my previous load development methods may be getting in the way of improving my skills…

SR is all worked out. Buy a barrel well machined from a good blank and load one of the standards then go work on yourself. That is what SR is all about. Mastery of yourself to effectively deploy a firearm more effectively than your fellow competitors firing nearly identical firearms. The IROC of the shooting world if you will.

When we talk of significant or practical difference one is based on being able to mathematically see a difference between two different samples (significant or not) and the other is based on experience and judgement. Hypothesis testing is a powerful tool if you understand how to deploy it but realistically it doesn’t hold much value in load development that most of us execute.