Follow up on reloading for ELR…

[boisepaw]

This is a follow-up to an earlier thread of mine. Remington 700 rifle in 308…reloading 185 grain Berger Juggernaut with Varget and H4895, Lapua brass, CCI primers. Shooting at 100 yards off bench, bipod and bags.

I’m trying to develop a load for that bullet. Last week I loaded 11 rounds with both H4895 and Varget at .2 grain increments to look for a ”node” of similar velocities. The Varget…loaded with 11 rounds from 43 to 45 grains (in .2 grain increments) yielded a group for all 11 rounds at different velocities…of 1.25 inches at 100 yards.

Yesterday was quite different.

The first photo is 42.9 grains of H4895. SIX rounds in 7/16 of an inch. Ave 2624 FPS, SD = 18, ES = 48. I love the group size but don’t like the SD and ES if stretching out to long range.

The second photo is 44.4 grains of Varget. Five rounds in 1.5 inches. Ave speed 2579 FPS, SD = 7, ES = 20. Lousy group (worse than with 11 rounds loaded at different charge weights) but acceptable SD and ES.

To add to the information/confusion, with 41.7 grains of H4895 I had another six shot group at ¾ inch but an SD of 13 and ES of 34.

Why the heck is such a GREAT group coming from such lousy SD and ES? And vice versa, such a modest group from a very nice SD and ES?

What gives?

 

[m1ajunkie]

What have you done with seating depth? I'd run a seating depth test with the varget and see what happens. Seating depth is something I've seen make a fairly noticeable difference on several of my loads recently.

ETA: I'm finding accuracy with much more jump than expected...... I plan to load up a 185 load for my SR25 when the bullets show up and I'm thinking I'll start at 2.80" OAL and this should shoot.

 

[MarshallDodge]

I tend to follow the rule of playing with powder charges to get a safe charge near the velocity I want and then play with seating depth to tighten groups. Depending on the ogive, some bullets are very sensitive to seating depth.

 

[Ledzep]

ES/SD and group size only show correlation in combinations that are absolutely awful for the given cartridge. For the vast majority of 'popular' combinations, there will be no correlation between group size and ES/SD, all else being as equal as possible.

Group size opening from velocity variation doesn't begin to show up until after 400-500yd.

 

[badassgunworks]

Standard deviation accuracy are not related

 

[MarshallDodge]

Standard deviation accuracy are not related

Agreed.

A low SD shows a good charge weight to create consistent velocity, it doesn't correlate to seating depth, bullet accuracy, etc.

 

[Ledzep]

Mean radius on the left at 200yd in inches. MV SD in fps on the bottom. This is 6mm ARC with 110 A-tips, each color a different powder in 3x 1.0gr increments, each dot 20 shots worth of data.

No real correlations to be found.

 

[boisepaw]

Mean radius on the left at 200yd in inches. MV SD in fps on the bottom. This is 6mm ARC with 110 A-tips, each color a different powder in 3x 1.0gr increments, each dot 20 shots worth of data.

No real correlations to be found.

View attachment 7890837

VERY interesting! Thanks.

 

[ptosis]

To start with, any conclusions based on a 5 shot group size are invalid, as the "data" is largely insignificant from statistical and engineering standpoints.

 

[badassgunworks]

To start with, any conclusions based on a 5 shot group size are invalid, as the "data" is largely insignificant from statistical and engineering standpoints.

So what would you suggest without doing some insane large sample test remember the entire industry for many many years has been doing testing based off of ladder fire shot groups

 

[ptosis]

So what would you suggest without doing some insane large sample test remember the entire industry for many many years has been doing testing based off of ladder fire shot groups

All I am saying is that for most practical purposes, as a measure of load precision, a single 5-shot group size is meaningless, the confidence interval is larger than the actual value.
(Caveat: such metrics can make sense to quickly discard totally bad loads, or totally bad barrels (the probability of a good load producing a bad group is smaller than of a bad load producing a good group), but then again -- you need to know your stats and understand the implications.)
To reiterate: as a mean of comparison of two reasonably decent loads, a 5-shot ES is not data, it is garbage.

 

[Taylorbok]

All I am saying is that for most practical purposes, as a measure of load precision, a single 5-shot group size is meaningless, the confidence interval is larger than the actual value.
(Caveat: such metrics can make sense to quickly discard totally bad loads, or totally bad barrels (the probability of a good load producing a bad group is smaller than of a bad load producing a good group), but then again -- you need to know your stats and understand the implications.)
To reiterate: as a mean of comparison of two reasonably decent loads, a 5-shot ES is not data, it is garbage.

What’s your alternative to the way winning competitors have been doing it for years?

 

[ptosis]

What’s your alternative to the way winning competitors have been doing it for years?

Google for "argumentum ad verecundiam".

Winning competitors? All? Most? Some? A few? One guy with whom my brother in law had a beer once? Claiming that "winning competitors" are evaluating the precision of a load on a basis of a single 5-shot group is… to stay polite… a bold statement.

I agree that an engineering degree is not required to be an excellent shooter. Given enough maniacal attention to reloading, the difference between two close loads is usually very small (unless there is a drastic qualitative factor in the game, such as one in the lands and not the other). People might prefer one load over another based on data, which is insignificant from the engineering standpoint, but it does not mean that the discarded one was bad, or even was worse. If they repeat the experiment, the results may be opposite.

Why do you think there isn’t a single army in the world measuring or qualifying firearms precision in terms of "5-shot group size"?

(Also, at long range knowledge of ballistics of your projectile, and the art of reading the wind, are immensely more important for the winning hit than "group size" at short range, but that’s an entirely different subject.)

 

[Sandow the Heretic]

Ok... I'm going to be the asshole and just say it. 308 ELR loading? Trying to push all the way out to 1000 yards? What is your goal?

The difference between your groups looks like a marginal stability issue. Too much bullet for your twist rate at that velocity. You try both of those loads at 800 yards and they will probably be equally terrible. High ES will have you high/low from shot to shot and the marginal stability will have you dancing all around.

If you are just pushing the limit of the cartridge, which is fun for sure, you need to find a load with a bullet that has a low ES AND a decent grouping. Maybe that is just not the bullet for your gun.

-Alex

 

[badassgunworks]

Google for "argumentum ad verecundiam".

Winning competitors? All? Most? Some? A few? One guy with whom my brother in law had a beer once? Claiming that "winning competitors" are evaluating the precision of a load on a basis of a single 5-shot group is… to stay polite… a bold statement.

I agree that an engineering degree is not required to be an excellent shooter. Given enough maniacal attention to reloading, the difference between two close loads is usually very small (unless there is a drastic qualitative factor in the game, such as one in the lands and not the other). People might prefer one load over another based on data, which is insignificant from the engineering standpoint, but it does not mean that the discarded one was bad, or even was worse. If they repeat the experiment, the results may be opposite.

Why do you think there isn’t a single army in the world measuring or qualifying firearms precision in terms of "5-shot group size"?

(Also, at long range knowledge of ballistics of your projectile, and the art of reading the wind, are immensely more important for the winning hit than "group size" at short range, but that’s an entirely different subject.)

I appreciate the engeneering view of things considering i am one. How ever i also dont
allow common sense to void of thoughts. Common pratice in all shooting venues is 5 shot ladder tests conforming the possibility of some form of consistancy in accuracy and standard deviations. A prosess that is repeatable confirms load quality . from a elr standpoint long shot strings conferms nothing other then the nature of the beast of generated heat from the use of large volumes of powder. It reflects nothing other then the circumstances. Yes wind is a huge factor but considering the unknown of the combined wind in flight path. its a arrogant claim by so many of the masters of elr . The wind readers when there was piles of pure luck involved from day to day. We are not seeing much consistency ( some yes ) in elr one week one person wins a match next match there in last place. Did they forget how to read wind from day to day or is there other variables involved ? One thing is for sure large bullets create more splash. higher bc bullets with consistancy and velocity off set some margens of error in wind skills. A bullet that stable threw transonic is more predictable. And accurate dope puts you close For a follow up.

 

[Ledzep]

Common pratice in all shooting venues is 5 shot ladder tests conforming the possibility of some form of consistancy in accuracy and standard deviations. A prosess that is repeatable confirms load quality .

Go repeat a bunch of 5 shot groups and see how consistent they are.

Group size, mean radius, ES/SD on MV, MPOI all walk around conducting the "same" test.

 

[badassgunworks]

Go repeat a bunch of 5 shot groups and see how consistent they are.

Group size, mean radius, ES/SD on MV, MPOI all walk around conducting the "same" test.

Do it all the time have been doing it for 30 years shooting bench rest with a 6mm ppc. I have dozens of rifles that at any time i grab one it will shoot 1 hole groups . At 100 yards i can shoot 10 5 shot groups sub .250 on any given day. Granted some will be in the .1's but it wont just randomly go to .500

 

[Ledzep]

So you repeat the same test and get different results?

Do you know the limitations of the variability of those tests?

Does that variability overlap with other loads?

How do you know if the single 5-shot string from a ladder you shot was the min, average, max, or anywhere in between them?

 

[badassgunworks]

If x devloped load ranges from group sizes ranging form .095 to .220 the avrage is some where between
But those variables are a combined variables of changes in environment and human factor but it doesn't turn into a 1/2 or 3/4 " group. so the devloped load ladder worked to determine what was good and what was not .

 

[Taylorbok]

@ptosis I understand your point, I do agree in the grand scheme of things 5 data points is very insignificant but Yes most winning competitors will initially evaluate a load off of ladder and then 5 or even 3 shot groups and obviously continue to test it's viability. I just assumed you had some proposition of an alternative route.

Military can afford the money, time, components and man power to do very rigorous testing, you can't compare that to a single shooter you'd never have time to actually shoot, you'd just be testing ammo all the time.

 

[badassgunworks]

@ptosis I understand your point, I do agree in the grand scheme of things 5 data points is very insignificant but Yes most winning competitors will initially evaluate a load off of ladder and then 5 or even 3 shot groups and obviously continue to test it's viability. I just assumed you had some proposition of an alternative route.

Military can afford the money, time, components and man power to do very rigorous testing, you can't compare that to a single shooter you'd never have time to actually shoot, you'd just be testing ammo all the time.

Even the guys who regurgitate comments from Applied Ballistics still preformed letter test there is no reasonable alternative

 

[Sandow the Heretic]

@ptosis I understand your point, I do agree in the grand scheme of things 5 data points is very insignificant but Yes most winning competitors will initially evaluate a load off of ladder and then 5 or even 3 shot groups and obviously continue to test it's viability. I just assumed you had some proposition of an alternative route.

Military can afford the money, time, components and man power to do very rigorous testing, you can't compare that to a single shooter you'd never have time to actually shoot, you'd just be testing ammo all the time.

People get hung up on sample size but the reality is that 3 is enough to call something crap. You can't get confidence in it being good with that few repetitions but you can certainly weed out the crap. The larger the differences are, the smaller the sample size needs to be to have statistical validity. If you have a 60 fps difference between two shots, you can't get a standard deviation from that but you can still call it a garbage load with decent confidence.

Fire sets of three to eliminate the worst loads and then try a larger sample of the remaining loads. 10 seems like good enough number to be reasonably confident of the results. Same endpoint with fewer shots fired.

-Alex

 

[badassgunworks]

So you repeat the same test and get different results?

Do you know the limitations of the variability of those tests?

Does that variability overlap with other loads?

How do you know if the single 5-shot string from a ladder you shot was the min, average, max, or anywhere in between them?

A knowledgeable person doing a ladder understands what is happening and what to expect during that ladder if during that test if somthing is out of the typical norm it can be repeated to conferm . Group size avrage is not part of why a ladder is performed good loads produce good groups on avrage bad loads produce bad groups on avrage.

 

[badassgunworks]

People get hung up on sample size but the reality is that 3 is enough to call something crap. You can't get confidence in it being good with that few repetitions but you can certainly weed out the crap. The larger the differences are, the smaller the sample size needs to be to have statistical validity. If you have a 60 fps difference between two shots, you can't get a standard deviation from that but you can still call it a garbage load with decent confidence.

Fire sets of three to eliminate the worst loads and then try a larger sample of the remaining loads. 10 seems like good enough number to be reasonably confident of the results. Same endpoint with fewer shots fired.

-Alex

You are 100% correct and 3 shots can also tell you if it is following a typical node wave. Thus telling you what might be a good place to further fine tune.

 

[badassgunworks]

Sample sizes only conferm what a person should already know burning out your throst is a ignorant endeavor to answer a useless question.

 

[Ledzep]

Node wave lol

You misspelled "White noise"

 

[Feniks Technologies]

Do it all the time have been doing it for 30 years shooting bench rest with a 6mm ppc. I have dozens of rifles that at any time i grab one it will shoot 1 hole groups . At 100 yards i can shoot 10 5 shot groups sub .250 on any given day. Granted some will be in the .1's but it wont just randomly go to .500

That only furthers his point.

Your example of 10 five shot groups and ranges from .100 to .250.

Which would mean the “noise” of the rifle and shooter is anything in between.


So, if you changed anything (load, tuner, seating, etc), it would have to shoot smaller than the .100 (and consistently) to be considered a change for the better. And would have to shoot worse than .250 to be considered worse.


Also, fwiw, I promise @Ledzep can show more recorded data than anyone who regularly posts on the forum.

 

[Ledzep]

Sample sizes only conferm what a person should already know burning out your throst is a ignorant endeavor to answer a useless question.

No, actually. Large sample repetitions of ladder tests show that nodes do not exist and are an artifact of small sample noise. Large sample tests show the variability of small sample tests and the repeatability of large sample tests. Large sample tests show realistic expectations of precision, MPOI, and ES/SD, and gives realistic inputs to evaluate LR and ELR hit probability.

ETA: There are three ways to approach load development. Two of them take into account the understanding of the information gleaned from large sample testing.

1) Load and go. Toss good components together, spot check it to make sure it fits your needs and carry on with life.
2) 20-30x shots of each variable, a statistically viable exhaustive approach
3) 1-shot, 3-shot or 5-shot ea. ladders, OCW, satterlee, etc... interpret scattered data, and fool yourself.

 

[Feniks Technologies]

What’s your alternative to the way winning competitors have been doing it for years?

The problem is that no one is collecting data to show this one way or another.

For example, no chronographs allowed at F class matches. So we have no idea which shots are running faster or slower.

Any long term data that is collected doesn’t support it. AB has let plenty come shoot to prove things like positive compensation and thus far, no one has been able to show them anything consistent.


Furthermore, no one ever takes the “bad” ammo and continues to test it. So, without further testing bad loads, you have nothing to compare it to.



The fact there are numerous “load development” methods is not actually a good thing. It’s like barrel break in. When you have many methods that work, but are so far off from one another, that typically indicates they don’t matter. As in, something works in spite of a method, not because of it.

Hell, look at the saterlee method. Hundreds, if not thousands of people “develop” loads with a method that can be disproven just running the test a few times over.

So, my rebuttal question, how do you explain top shooters (and there are some in F class) that use methods such as this to win?

For methods to be valid, you’d need to see quite a few that don’t work. Which would drive everyone who wanted to be a top contender to use those methods if they want to win.

You will see trends in equipment because that’s what it takes to win. You should also see trends in load development methods if they are valid.

 

[Taylorbok]

@Feniks Technologies @Ledzep
.100 to .250 being in the noise of the rifle and shooter is IMO especially for ELR well within the needs of the ammo and shooter.
I used to believe in nodes and sort of still do believe it's a viable way to track down a load that will produce a low ES/SD, the only thing I have to support that is my own data and several rifles that shoot ES under 15 across 10 shots. (decent, not unreal)

That being said I needed to back a load down the other day because I was going to be shooting in much hotter weather. I went back to my old ladder test data and the next node down was slower than I wanted to be. (would have been fine but hey shoot'em fast). I literally picked the charge that was about the speed I wanted to be (no where near a "node") loaded up 5 at my old seating depth, shot .250" and ES of 9....

 

[Feniks Technologies]

@Feniks Technologies @Ledzep
.100 to .250 being in the noise of the rifle and shooter is IMO especially for ELR well within the needs of the ammo and shooter.
I used to believe in nodes and sort of still do believe it's a viable way to track down a load that will produce a low ES/SD, the only thing I have to support that is my own data and several rifles that shoot ES under 15 across 10 shots. (decent, not unreal)

That being said I needed to back a load down the other day because I was going to be shooting in much hotter weather. I went back to my old ladder test data and the next node down was slower than I wanted to be. (would have been fine but hey shoot'em fast). I literally picked the charge that was about the speed I wanted to be (no where near a "node") loaded up 5 at my old seating depth, shot .250" and ES of 9....

He used a 6ppc as an example.

You likely won’t see a ton of people shooting large cartridges that small. The point being, it’s all relative.


Also worth nothing for you to have an true ES of 15, you’d need a sub 4sd. If your 10 shot groups aren’t show under a 4sd, that’s not your true ES.

And even so, if you shot a 10 shot string with say a 4sd, your margin of error could go as high as around 7sd and that would be a 28 es most of the time.

So, for you to have an absolutely true 15es with your 10 shot string you would have to have a 2sd *max*.


That’s how all this long term dispersion stuff works.

So, if you’re not seeing a sub 2sd every 10 shot string, you do *not* have a long term es of 15.

 

[ptosis]

+1
ES on small samples is an extremely poor measure, as it is too much subject to random noise.
SD is way, way better, but then again, in order to have something meaningful, you need a certain number of shots anyway.

Here:

rows = SD in m/s (to convert to archaic units, multiply by 3.3)
columns = number of shots
values = 95% confidence interval (+/-) of the values obtained

Example: on a string of 10 shots, SD is 3 m/s (roughly 10 fps). Well, the confidence interval is [0.9 .. 5.1] m/s, that is ~ 3 .. 17 fps.

 

[badassgunworks]

He used a 6ppc as an example.

You likely won’t see a ton of people shooting large cartridges that small. The point being, it’s all relative.


Also worth nothing for you to have an true ES of 15, you’d need a sub 4sd. If your 10 shot groups aren’t show under a 4sd, that’s not your true ES.

And even so, if you shot a 10 shot string with say a 4sd, your margin of error could go as high as around 7sd and that would be a 28 es most of the time.

So, for you to have an absolutely true 15es with your 10 shot string you would have to have a 2sd *max*.


That’s how all this long term dispersion stuff works.

So, if you’re not seeing a sub 2sd every 10 shot string, you do *not* have a long term es of 15.

With the 408 and 416 Colossus our average group size is around.500 standard deviations was Final developed load has been an average of 2 feet per second on five shot strings I had several five shot strings that were 0

 

[Feniks Technologies]

Here’s a screenshot of a post I made in a few platforms.

It gives you the SD required vs the sample size to have a known true single digit SD in the long run.

Obviously there will be some exceptions. But this will cover roughly 95%.

 

[Feniks Technologies]

With the 408 and 416 Colossus our average group size is around.500 standard deviations was Final developed load has been an average of 2 feet per second on five shot strings I had several five shot strings that were 0

So, a 2sd on a 5 shot string:

Could be as high as 5.7sd which would be around a 24 es 95% of the time. Obviously that is the highest. It could be as low as just over 1fps sd which would be under 5es 95% of the time.

 

[badassgunworks]

Obviously the more you shoot the more the possibility of variation there are many factors involved other than just pure chance like for example Heat. However a lot of that is only relevant in relation to how many shots you're taking in a specific venue. The question of repeatability for me is today I take three shots tomorrow I take three shots and the day after that I take three shots all coming up extremely close to one another and this was my point earlier on any given day what size of a group can you shoot randomly with your rifle

 

[badassgunworks]

So, a 2sd on a 5 shot string:

Could be as high as 5.7sd which would be around a 24 es 95% of the time. Obviously that is the highest. It could be as low as just over 1fps sd which would be under 5es 95% of the time.

on the last test I did I did not have an extreme spread that was four feet per second. And that was last week

 

[Taylorbok]

@Feniks Technologies
I totally get the long term dispersion, as stated the more you shoot the more variance you will have. Average SD is 5 according to Labradar, It's best I have to go by. I don't chrono every round usually just during testing and then a few strings for confirmation. Been good enough for me so far (longest impacts 3/5 at 2189 yards with 300 RUM). Still looking to stretch it out next goals of 2500, 3000, 3500

We've successfully completely derailed the thread, back to scheduled programming.
@boisepaw if you are looking for ELR performance of that 308 I would take the lowest sd/es charge you have and do a seating depth test until it groups how you want it.

 

[Sandow the Heretic]

+1
ES on small samples is an extremely poor measure, as it is too much subject to random noise.
SD is way, way better, but then again, in order to have something meaningful, you need a certain number of shots anyway.

Here:
View attachment 7895495

rows = SD in m/s (to convert to archaic units, multiply by 3.3)
columns = number of shots
values = 95% confidence interval (+/-) of the values obtained

Example: on a string of 10 shots, SD is 3 m/s (roughly 10 fps). Well, the confidence interval is [0.9 .. 5.1] m/s, that is ~ 3 .. 17 fps.

So if this is real data then I'd point out that the three shot number is awfully predictive of the 25 shot number.

-Alex

 

[Ledzep]

I have some 100-shot data sets for factory ammo. These 100 shots are then broken (sequentially) into smaller groups. NO variables being tested other than perception based on sample size. These are shot from a 1.250" straight accuracy test barrel in a pneumatic/hydraulic return rail gun in an indoor 200yd range.

Group size average:
100 shot: 1.12 MOA
Avg of 3x 33 shot: .959 MOA
Avg of 5x 20 shot: .829 MOA
Avg of 10x 10 shot: .738 MOA
Avg of 20x 5 shot: .569 MOA
Avg of 33x 3 shot: .477 MOA


Group size variation:
33-shot: .872 to 1.12 MOA
20-shot: .586 to 1.11 MOA
10-shot: .4875 to 1.05 MOA
5-shot: .273 to .793 MOA
3-shot: .133 to .750 MOA

There is a natural tendency for smaller sample size to produce smaller spreads (perceptibly better results) and more variability test-to-test. Measuring mean radius produces less variability than group size, but it still takes 20-30 rounds to really settle out. 18 of the 33, 3-shot groups were less than the average. So odds are in your favor to produce very small 3-shot groups from what is basically "okay" factory match ammo. Odds are very small for you to encompass the worst-case dispersion of the system with 3 random shots.

 

[ptosis]

For the geeks out there: http://ballistipedia.com/
(caution: lvl. 80 maniacal nitpicks)

And my [rather awkward] practical implementation: https://bc.geladen.ch/taran/taran.html

To confirm @Ledzep 's findings, here's what the science says

Size of the group, expressed as multiple of Rayleigh's distribution standard deviation

Average max spread (3-shot)	2.41
Average max spread (5-shot)	3.06
Average max spread (10-shot)	3.79
R50 (50% hit radius)	1.18
R95	2.45
R99	3.03

 

[ptosis]

So if this is real data then I'd point out that the three shot number is awfully predictive of the 25 shot number.

-Alex

I'm not sure: either I did not describe it clearly enough, or you are interpreting it the way I did not expect. What exactly do you mean?

 

[Sandow the Heretic]

I'm not sure: either I did not describe it clearly enough, or you are interpreting it the way I did not expect. What exactly do you mean?

Looking at the data it is too perfectly linear to be actual chronological data so this is simply supposed to be illustrative. What I was pointing out was that at 3 shots the relationship to the CI at 25 shots was highly correlated. Looking at Ledzep's data though, his is pretty linear as well. The shoes back to my earlier point that very large sample sizes are unnecessary.

That said, if you have a magneto speed on your gun at all times, why not collect data for analytics.

-Alex

 

[badassgunworks]

I have some 100-shot data sets for factory ammo. These 100 shots are then broken (sequentially) into smaller groups. NO variables being tested other than perception based on sample size. These are shot from a 1.250" straight accuracy test barrel in a pneumatic/hydraulic return rail gun in an indoor 200yd range.

Group size average:
100 shot: 1.12 MOA
Avg of 3x 33 shot: .959 MOA
Avg of 5x 20 shot: .829 MOA
Avg of 10x 10 shot: .738 MOA
Avg of 20x 5 shot: .569 MOA
Avg of 33x 3 shot: .477 MOA


Group size variation:
33-shot: .872 to 1.12 MOA
20-shot: .586 to 1.11 MOA
10-shot: .4875 to 1.05 MOA
5-shot: .273 to .793 MOA
3-shot: .133 to .750 MOA

There is a natural tendency for smaller sample size to produce smaller spreads (perceptibly better results) and more variability test-to-test. Measuring mean radius produces less variability than group size, but it still takes 20-30 rounds to really settle out. 18 of the 33, 3-shot groups were less than the average. So odds are in your favor to produce very small 3-shot groups from what is basically "okay" factory match ammo. Odds are very small for you to encompass the worst-case dispersion of the system with 3 random shots.

So is this example a shity shooter? A shity gun? Or crappy loads?

 

[ptosis]

Looking at the data it is too perfectly linear to be actual chronological data so this is simply supposed to be illustrative. What I was pointing out was that at 3 shots the relationship to the CI at 25 shots was highly correlated. Looking at Ledzep's data though, his is pretty linear as well. The shoes back to my earlier point that very large sample sizes are unnecessary.

That said, if you have a magneto speed on your gun at all times, why not collect data for analytics.

-Alex

This is by far not linear. As you can see, with the increasing number of shots (top-down) for a given SD (column), the confidence interval does not decrease linearly; it very quickly gets to "diminishing returns".

And -- just to clarify -- the values in the table represent confidence intervals, that is -- the certainty of the results. For example, if we measure an SD value of 4 with a CI of +/-3, that means that [simplifying] the real value can be expected somewhere between 1 and 7.

And -- no, I do not permanently run a chrono on or near my barrel, because most of the time I just don't care. The table represents confidence interval for a population standard deviation, unknown mean, for normal distribution -- a universally accepted model for this type of things in ballistics research.

 

[Ledzep]

Looking at the data it is too perfectly linear to be actual chronological data so this is simply supposed to be illustrative. What I was pointing out was that at 3 shots the relationship to the CI at 25 shots was highly correlated. Looking at Ledzep's data though, his is pretty linear as well. The shoes back to my earlier point that very large sample sizes are unnecessary.

That said, if you have a magneto speed on your gun at all times, why not collect data for analytics.

-Alex

Do not confuse the averages of a BUNCH of small sample tests with the results of a single small sample test. That's the whole point we're trying to make. When you shoot a single 3-shot group you are at random odds to get anywhere between .133 to .750 MOA (using my data as an example). There is no scalar factor that can be applied to that single group to accurately determine the total dispersion pattern of a larger sample test. There are scalar factors that can be applied to the AVERAGES of a bunch of tests, but that's not the same. To get enough data to apply those scalar factors you basically might as well shoot a single large sample test and correlate POA and POI for all of the shots.

So is this example a shity shooter? A shity gun? Or crappy loads?

It's in the first line(s) of the post. Factory match ammo in a Bartlein 1.250" barreled rail gun with a pneumatic trigger. No shooter error. It's 1 MOA factory ammo.... Or 1/2 MOA factory ammo if you shoot 3 shots.

 

[Taylorbok]

@Ledzep that's some super interesting data. I'd love to see that same test shot with a hand load that someone competent has tuned.

 

[Ledzep]

I've ruined benchrest for more than one person. Guy thinks he's in the .1's-.3's with a 15fps ES and a 20-30 shot set will show otherwise.

I have some handload data, 20 and 35 shot ladder tests with a couple of different cartridges. Somewhere in the .4-.6 MOA range is about as good as I've ever seen with SD's in the 5-7fps range being the best that are repeatable. The powder ladder is less important than the powder type.

 

[badassgunworks]

I've ruined benchrest for more than one person. Guy thinks he's in the .1's-.3's with a 15fps ES and a 20-30 shot set will show otherwise.

I have some handload data, 20 and 35 shot ladder tests with a couple of different cartridges. Somewhere in the .4-.6 MOA range is about as good as I've ever seen with SD's in the 5-7fps range being the best that are repeatable. The powder ladder is less important than the powder type.

Thats the problem you cant stay in the .1s and .3s having ES of 15

 

[Feniks Technologies]

Thats the problem you cant stay in the .1s and .3s having ES of 15

So which is it, in your opinion? Is SD and accuracy related or not?