Variations in SD

#2
There is no “required” with statistics, you just want a statistically significant population to sample from. If 3 shots is your standard that’s fine, I say 10 is better if you are wanting a hard number to plug into your calculator, 30 is better and you won’t get single digit sd with that many shots.
 
#3
^^^what he said

With 20-30 rounds making up your sd You won’t get a tiny sd but you’ll get closer to your TRUE sd. And if you get close to your true sd, you can expect it to be fairly consistent string to string.

Shooting 3 or 5 rounds you might get lots of small SDs but they won’t be consistent.
 
  • Like
Reactions: R0man
#4
Also note that SD and group size do not necessarily correlate Depending on the range, you can quite a large SD with small groups and large groups with a small SD.

If you do a search for confidence level calculator, you'll see several ways to determine the relative accuracy of various sample sizes. For example, given a sample size of 5, an SD of 6 with a 95% confidence level in the measurement the actual SD may vary by +/- 7.46, where increasing teh sample to 20 decreases the error level to +/- 2.8.

Try reading these articles for details.

Shooters statistics
 
#6
Umm, you cannot test much of anything with a sample of two. In point of fact , SD cannot be computed with a sample of two. ES is a pretty useless number, as in any population it's a sample of two.
 
#7
Five shots during the initial OCW (may move to three shots going forward due to supplies scarcity) then once I think I have a load I’ll do a ten-shot string, once in the summer and once in the winter, re-taking measurements if I switch to a different lot of powder due to me exhausting my supply of the prev lot.
 
#8
CoryT said:
Umm, you cannot test much of anything with a sample of two. In point of fact , SD cannot be computed with a sample of two. ES is a pretty useless number, as in any population it's a sample of two.
Click to expand...
You can do very much of figuring out where you don’t want to be with it.

It may not be able to give you an answer to four decimal places but you can tell if a value is positive or negative.
 
#9
nn8734 said:
Five shots during the initial OCW (may move to three shots going forward due to supplies scarcity) then once I think I have a load I’ll do a ten-shot string, once in the summer and once in the winter, re-taking measurements if I switch to a different lot of powder due to me exhausting my supply of the prev lot.
Click to expand...

The OCW test is not testing velocity, it's testing vertical spread. The population for that test is normally 12-18 rounds, with the datapoint of interest covering at least six of those. Ten rounds for velocity mean is OK, but it will not give you a very accurate SD.
 
#10
CoryT said:
The OCW test is not testing velocity, it's testing vertical spread. The population for that test is normally 12-18 rounds, with the datapoint of interest covering at least six of those. Ten rounds for velocity mean is OK, but it will not give you a very accurate SD.
Click to expand...

right, yea I know that however I’m also taking MV data at the time of the OCW. I then validate with 10 shots once I think I have a good load.

I’ve never had an issue where 10 rounds wasn’t sufficient to tell me what I needed to know about a load. But if you’re referring to small sample testing of a population to make inferences about that parent population, minimum sample size requirements is 30 if I’m not mistaken.
 
#11
spife7980 said:
You can do very much of figuring out where you don’t want to be with it.

It may not be able to give you an answer to four decimal places but you can tell if a value is positive or negative.
Click to expand...

Lets suppose you have an actual SD of 10 FPS. You fire two shots. The ES is 60 fps. Is that cause to reject the load? No, because those two shots fall withing the 95 percentile of all shots you will fire with that load, +/- 3 SD's. What exactly do you expect to figure out with two shots? We could tell it's over pressure. We could see it's not stable in the twist rate if we see keyholing. If they are 5" apart at 100 and we can eliminate any other possible errors you could perhaps reject it as a load. You might decide the powder is too fast or slow based on the velocity, but assumes you are going 'off book' rather than using an established loading base. That's about all you might be able to discern.
 
#13
nn8734 said:
right, yea I know that however I’m also taking MV data at the time of the OCW. I then validate with 10 shots once I think I have a good load.

I’ve never had an issue where 10 rounds wasn’t sufficient to tell me what I needed to know about a load. But if you’re referring to small sample testing of a population to make inferences about that parent population, minimum sample size requirements is 30 if I’m not mistaken.
Click to expand...

It depends on what you need to know. The requirements for ELR are not the same as for PRS, neither of which are same for hunting. How accurate do you need to be for the purpose at hand? The Applied Ballistics WEZ calculator is a neat tool to find the parameters that actually matter. Obsessing over a single digit SD is probably not actually a solution to hitting 2 MOA targets in a PRS match. However I'd actually take .5 MOA groups at 100 with an SD of 6 over .3 MOA groups with an SD of 12 if I plan on shooting at 2K meters.
 
#14
CoryT said:
It depends on what you need to know. The requirements for ELR are not the same as for PRS, neither of which are same for hunting. How accurate do you need to be for the purpose at hand? The Applied Ballistics WEZ calculator is a neat tool to find the parameters that actually matter. Obsessing over a single digit SD is probably not actually a solution to hitting 2 MOA targets in a PRS match. However I'd actually take .5 MOA groups at 100 with an SD of 6 over .3 MOA groups with an SD of 12 if I plan on shooting at 2K meters.
Click to expand...

Agree in that i don’t obsess over single digit SD, it’a one of many parameters by which to judge a load and yes, use case matters. ELR, everything is more exacting including what you are willing to tolerate insofar as deviation from the mean and extreme spread of your rounds. For example, I tested Federal’s 130 Berger 6.5 creedmoor load twice, 10 random rounds each time. Avg SD was 17.4. I had no problem striking 2/3 ipsc steel targets out to 1100 yards with that stuff in reasonable conditions. SD is only part of the equation; for example, a load’s stability across varying conditions is more important to me than simply chasing low SD.
 
Last edited:
#15
spife7980 >I figure out that I don’t want a kid that presents 60 es ever.

Except as I just explained, an ES of 60 is perfectly possible with an SD of 10, which is pretty good. In fact, the ES could be 30 and your actual SD could be 5, in a sample of 20 rounds. Do you know what SD actually tells you? Do you understand a bell curve of normal distribution?
 
Last edited:
#18
spife7980 > I know that I don’t want to shoot an es of 60 so I would say it’s fine it’s job very well and very simply.

Why not? An ES of 60 correlates to an SD of 10, a value that many people strive for.
 
#20
CoryT said:
Umm, you cannot test much of anything with a sample of two. In point of fact , SD cannot be computed with a sample of two. ES is a pretty useless number, as in any population it's a sample of two.
Click to expand...

I’d contest ES is not useless. If I run 20 (or 50), whatever, shots over the chrono, and I have say an ES of 30 and an sd of 10, I know that all my shots will be within the 30 and no chance of opening up shots larger than my target.

If I do this and I have an ES of 60 and an sd of 6, then some of my shots will actually miss my target (assuming the target is smaller than the ES allows for).

You can have a high ES and a low SD. But you’ll never have a low ES and a high SD.

While yes, the ES is two shots, it contains *all* others within that threshold.
 
#21
I typed a couple things and deleted to save my breath. Practical methods will never satisfactory some people when there is a decimal involved and they think they can math the results to their benefit.


So in short: I don’t give a shit what the math says it could theoretically be, I don’t want 60 es ever so I move on. It doesn’t take another 20 bucks in ammo and numbers and being good will hunting to figure that out.
 
#22
spife7980 said:
I typed a couple things and deleted to save my breath. Practical methods will never satisfactory some people when there is a decimal involved and they think they can math the results to their benefit.


So in short: I don’t give a shit what the math says it could theoretically be, I don’t want 60 es ever so I move on. It doesn’t take another 20 bucks in ammo and numbers and being good will hunting to figure that out.
Click to expand...

I have run thousands of rounds over my chrono. So I agree.

I have used chrono and such to make sure my loading process (when I want to be picky) is extremely consistent. When I use the pickier methods, my 5 shot strings give me a very good idea what my 20 and 50 shot strings will look like.

So, I use this info gathered over time to use smaller sample sizes to make decisions. At least which ones to throw in the garbage.
 
  • Like
Reactions: nn8734 and spife7980
#23
Dthomas3523 said:
I’d contest ES is not useless. If I run 20 (or 50), whatever, shots over the chrono, and I have say an ES of 30 and an sd of 10, I know that all my shots will be within the 30 and no chance of opening up shots larger than my target.

If I do this and I have an ES of 60 and an sd of 6, then some of my shots will actually miss my target (assuming the target is smaller than the ES allows for).

You can have a high ES and a low SD. But you’ll never have a low ES and a high SD.

While yes, the ES is two shots, it contains *all* others within that threshold.
Click to expand...

Except that's not how it works. Outliers are a problem when you analyze a data set. Having a string of 20 shots gives you a pretty reliable SD. If that SD is 6, then 99% of the shots you fire should be within +/- 18 FPS, creating a 36 FPS ES possible over that 99% set. It does not eliminate the possibility of +/- 24 FPS shots, or a 48 FPS ES. It's just that over the course of say 100 shots that's only likely to happen once.

If you have an SD of 10, an ES of 60 will happen some 5% of the time, with a larger ES possible. Given a data set of two shots, the ES tells me nothing at all, unless it's something silly like 150 FPS. A value of lets say 50 is basically meaningless. It's entirely possible that the actual SD of that load is 10, and the next 98 shots will have an ES of 40, and 68 of them will be in an ES of 20. It's also entirely possible to fire two shots with an ES of 0 and have a load where the SD is 30. The data set is too small to say it's not possible. Given 100 shots of almost any load you have you will almost certainly find two at the exact same velocity. It's not surprising to find them consecutively.

Outliers in the data set happen. Maybe it was a bad primer, tight case neck, light bullet, lube left on the case... Nothing prohibits that from being in the two rounds you test.

Take your own example, an SD of 6 and a large (20-50) sample where the ES is 60. That's not impossible, simply unexpected. If you eliminated those two shots, the SD would shrink quite a bit, and you'd think the load was some kind of miracle. If even one of the two was removed, you'd be very happy.

In the end, chronograph numbers are not really what we care about anyway. Given enough data, it can perhaps tell us something, but I'd venture to say most people don't know, nor do they care what that something is. The majority of handloaders could do without the chrono, just shoot a few groups at the longest range they normally use and they would know what they need to know. If I shoot 5 5 round groups at 800y and they are all under 8" and the MPI is within .1 mil of the POI, do I really care if the SD is 6 (Cool!) or 17 (Gag)? If I miss a 12" plate next week at the match, what's the most likely problem? It's not that that round was faster than average by 35 FPS.
 
#24
spife7980 said:
I typed a couple things and deleted to save my breath. Practical methods will never satisfactory some people when there is a decimal involved and they think they can math the results to their benefit.


So in short: I don’t give a shit what the math says it could theoretically be, I don’t want 60 es ever so I move on. It doesn’t take another 20 bucks in ammo and numbers and being good will hunting to figure that out.
Click to expand...

You already have loads that have an ES of 60, you just don't shoot with a chronograph every time so you don't see it when it happens. The math is what it is, no matter what you want or if you care.

I don't expect that you would say a rifle shoots 1/4 MOA because you fired 2 shots that overlap. There are reasons to shoot 5 rounds for a group and then shoot multiple groups before one states anything about the precision of a system.

However, the data can only tell you so much. SD for example, as stated, has no direct relation to accuracy or precision. Think of the chronograph like a dynomometer for a race car. Nice to know the horsepower, but what does that tell me about how the car corners?

I suspect most handloaders would actually be happier if they just left the chronograph at home.
 
#25
CoryT said:
Also note that SD and group size do not necessarily correlate Depending on the range, you can quite a large SD with small groups and large groups with a small SD.

If you do a search for confidence level calculator, you'll see several ways to determine the relative accuracy of various sample sizes. For example, given a sample size of 5, an SD of 6 with a 95% confidence level in the measurement the actual SD may vary by +/- 7.46, where increasing teh sample to 20 decreases the error level to +/- 2.8.

Try reading these articles for details.

Shooters statistics
Click to expand...
Thanks for your insight this has definitely helped make sense of the results I have been getting. I Chrono every shot and have recorded multiple 5 shot groups achieving SDs as low as 2 and as high as 14 with same load over and extended time period. As such I have recorded the average from batches of 25 shots as my SD figure in my ballistic calculator.
 
#26
Using all 25 rounds should give you a confidence interval of about 92% that the true SD is within 25% of the calculated value. If the SD of the 25 shot population is 10, the actual value should be between 12.7 and 7.5 That's rough, I did not plug everything into a confidence level calculator. You should be within a few FPS of the actual mean velocity. Plenty good for most any usage, if the group sizes are also good.
 
#27
CoryT said:
Using all 25 rounds should give you a confidence interval of about 92% that the true SD is within 25% of the calculated value. If the SD of the 25 shot population is 10, the actual value should be between 12.7 and 7.5 That's rough, I did not plug everything into a confidence level calculator. You should be within a few FPS of the actual mean velocity. Plenty good for most any usage, if the group sizes are also good.
Click to expand...
Thanks I will continue to monitor and refine, getting comfortable with shooting a new rifle (264 WM) group sizes are getting there .50 -.75 MOA 5 shot groups and .25 -.50 3 shot groups.
 
#28
3 shot groups don't count. Take multiple 5 shot groups and overlay them to get the true distribution of your shots. On Target TDS or Ballistic X can help with data collection and storage.

Of course as I'm fond of saying to the clients in class, if you want a small group, shoot once.
 
#29
Just do what the Jersey shooters do. Shoot 3 shots through your chronograph until you get sub 5 SD, and post away on social media. You'll impress all the n00b shooters who don't understand a thing about statistics, and become a shooting hero in the social media world.

If you can't get 3 shots to produce the results you want, or it's taking too long, just selectively delete shots from the string until you get the results you want.
 
  • Like
Reactions: Timo Turl
#30
kthomas said:
Just do what the Jersey shooters do. Shoot 3 shots through your chronograph until you get sub 5 SD, and post away on social media. You'll impress all the n00b shooters who don't understand a thing about statistics, and become a shooting hero in the social media world.

If you can't get 3 shots to produce the results you want, or it's taking too long, just selectively delete shots from the string until you get the results you want.
Click to expand...
i find it takes about 7 shots to get a 3 shot group with an SD of 1-2. 50% of the time every time
 
#31
kthomas said:
Just do what the Jersey shooters do. Shoot 3 shots through your chronograph until you get sub 5 SD, and post away on social media. You'll impress all the n00b shooters who don't understand a thing about statistics, and become a shooting hero in the social media world.

If you can't get 3 shots to produce the results you want, or it's taking too long, just selectively delete shots from the string until you get the results you want.
Click to expand...

Hence my “load development doesn’t matter for prs” mentality.

The fact that everyone (including the winners) do all kinds of random stuff snd post 3 shot strings, only furthers my point.
 
You must log in or register to reply here.

Similar threads

D
6.5 Creedmoor 100gr ELD-VT + Varget Load Development
Replies
31
Views
5K
Bevo
Bevo
B
Producing Ammo with Low SD's
Replies
56
Views
2K
V-Ref
V-Ref
NORCAL50
  • Poll
Which small rifle primer consistently gets you single digit SD's in 6mm+ cartridges?
Replies
17
Views
1K
Cody S
Cody S
B
Garmin Chronograph - calibration?
Replies
19
Views
1K
Baron23
Baron23
T
Super low SD but crappy groups, what to do
Replies
20
Views
905
OkieMike
O
Top Bottom
Back
Home
Forums
Merch
Alerts