How important is Standard Deviation?

Standard deviation based on a sample of only 5 or so is next to meaningless. Not completely meaningless, but knocking on the door.

It's meaningless if it's substantially different than SD for 50 shots. We simply use 5 because it's less costly. If the SD for 5 shots is 50, that is quite meaningful indeed. It means something is causing some meaningful inconsistency. I think denton's point is that if the SD for 5 shots was 10, it would be little guarantee that the results would remain consistent for the 6th shot. That's true. But if we're seeing inconsistency with 5 shots, we will rightfully have a desire to seek better consistency.

An SD of 50 is not proof of anything awful, and you're right to suspect that it could coincide with results that are nevertheless good at shorter ranges where wider variation in velocities makes less of a difference in point of impact than it would at longer ranges.

I don't have a specific suggestion based on experience with that cartridge, but I would suggest trying to change one variable at time. Try a different primer. If you don't see an improvement, go back to the first primer and try a different seating depth. Record what the result. Full-length resize and see if you get more consistency after that. If you don't try, you'll never know.

With that bullet, at that speed, a 50fps difference is 1" at 500 yards
Go with the smallest group size and walk away....

A slight breeze is going to give a 40 gr. bullet more issues at 500 yds. than a high SD ever will

Let me try that again.

When you measure 5 shots, you are taking a sample of the large population of shots that might come out of the rifle.

If you have a load that truly has an SD of 50, and if you keep on drawing many samples of 5, 95% of those samples will fall between about 25 and 120. Samples of 5 cannot resolve differences any finer than that, so you don't know where in that range you are. On top of that, SD has a built in bias in small samples.

So, for purposes of determining real population SD, a 5 shot sample is next to useless. It does not have a strong ability to predict the SD of the next 5 shot sample.

If you want to slow it down, try a heavier bullet.

I have good luck with ladder tests. Google will help you here.

Maybe try walking your targets out further. This won’t help your SD, but you can get a better idea what your groups look like at distance. For me it would give me a better idea what I’m dealing with at realistic hunting ranges. Plus, it makes you really concentrate on your shooting as mistakes are magnified. As mentioned above, a soft fart will blow a 40gr bullet around though.

I am of the not enough data camp. I suggest you take the most accurate group from your ladder test and load 20 of them. That will give you enough data, and will tell you more about accuracy as well. An SD of 50 speaks to rather extreme spreads which isn't usually consistent with accuracy.

denton said:

So, for purposes of determining real population SD, a 5 shot sample is next to useless. It does not have a strong ability to predict the SD of the next 5 shot sample.

Click to expand...

I completely agree with this, but not that a 5-shot SD is at all useless -- of course, you didn't state that it was, just that it was practically useless for determining real population SD which is true. So where I think 5-shot SD is useful, is when it reveals a problem in consistency. If I see too great a standard deviation in 5 shots, I don't need to know more about the real population to know that I have a problem. If I shoot just two shots and one is 4000fps and the next one is 4100fps, I have a std dev of 50 and I know there is a problem.

colton1337 said:

When performing a ladder test on 6 different loads I noticed that my standard deviation would not drop below the 50's.

Click to expand...

When I first got my chrono I couldn’t believe how large the SDs were for loads I had worked up that shot really small groups. When they say holes on paper is everything, they mean it.
While the PR crowd wants SDs in the single digits, their vertical drop at 1K is pretty important to predict reliably and consistently. My .22-250 load’s SD is in the 30’s and at the 100-500 yd range, between the trigger, wind and hold, seems to work just fine.
What was your ES? For rifle I look at thAt first, even in the small sample size. I throw those numbers in the ballistics app and see what that does to the trajectory at the ranges I’m interested in and then decide if I want to try to chase reducing ES.
There’s a lot of variables involved, and case prep is probably one of the most crucial. The other thing I try to shoot for is hitting a velocity node if it happens to be an accuracy node. Good luck!

Some specs that might be important are...

• 40gr. Sierra Blitzking
• 39.6gr. H380
• Federal 210M primers
• Winchester brass trimmed to ~1.902 in and neck sized only
• 2.390 OAL ( Can't remember exactly how far I am from the rifling)
• 1.975 to Ogive
• Target was at 100 yards

Click to expand...

Then throw in average, ES, SD and total number of shots. At 100 yards SD is pretty far down on the list of importance. As illustrated by your ragged hole 100 yard group with an SD greater than 50 FPS.

denton said:

Standard deviation based on a sample of only 5 or so is next to meaningless. Not completely meaningless, but knocking on the door.

Click to expand...

The 5 shot group is the starting point for future load development. I use it to find a starting point to move to the next gate in load development. If the 5 shot spread shows promise in SD and accuracy then I’ll load larger groups with that bullet/powder combination and run those through the chronograph. O.A.L. Is another part of load development that is a separate conversation.

It’s a process some take with load development while others stop at a reasonably accurate load.

MEHavey said:

With that bullet, at that speed, a 50fps difference is 1" at 500 yards
Go with the smallest group size and walk away....

Click to expand...

An SD of 50fps does not equal a variation of +/- 50fps. It means that ~68% of the shots will be in that range IF there is a normal distribution. ~95% will fall within +/- 100fps. With 5 shots you can't even say if it is a normal distribution, so those predictions aren't necessarily valid.

I know statistics, but I don't have any experience putting a large sample of shots over a chrono, so I can't say that there should be a normal distribution. But I can say, any prediction based on an SD assumes a normal distribution.

You didn't post the other numbers and each shot fps?
What is the ES, average etc.?

Regardless, look at the target, does obsessing about a number really matter?? Looks pretty tight to me??

An SD of 50fps does not equal a variation of +/- 50fps.
It means that ~68% of the shots will be in that range.

Click to expand...

I know that.

But rather than getting caught up in the esoterics of statistics, we just lay out the "who cares...?" factor.
Call it 3σ (± 150fps) to collect the kitchen sink, and you're still only ±3" -- minute of ground hog at half a kilometre.

At that point "who cares?"
Go for smallest group
(notice I didn't say "accuracy")

MEHavey said:

I know that.

But rather than getting caught up in the esoterics of statistics, we just lay out the "who cares...?" factor.
Call it 3σ (± 150fps) to collect the kitchen sink, and you're still only ±3" -- minute of ground hog at half a kilometre.

At that point "who cares?"
Go for smallest group
(notice I didn't say "accuracy")

Click to expand...

I bet the ground hog cares. :O

It's meaningless if it's substantially different than SD for 50 shots. We simply use 5 because it's less costly.

Click to expand...

Bingo. Your basically taking a sample. Any higher sample will be more statistically viable, obviously. There’s no argument there. I know of zero people that make up 50 rounds batches to test on their velocities and the corresponding stats. That would be expensive and time consuming indeed. IMO it’s not needed either.

At that point "who cares?"
Go for smallest group

Click to expand...

Sometimes this can be misleading. When I shoot Bergers out of my 300 WM, they start stabilizing after 100 yards. So, if I chase groups at 100, I have different results at further distances. I built the rifle and loads for long distance. My groups are adequate at 100 yards to drop a deer or an elk easily. My SDs are single digits.

Paper results will always trumps any crony numbers. If you happen to find the right load for the harmonics in your barrel it auto compensates for small variations.

...Bergers out of my 300 WM, they start stabilizing after 100 yards.

Click to expand...

Someone smarter than me come back with the general physics how a passive object
-- once launched -- can reduce its dispersion angle over longer flight distance.
(It'll give me hope.... )


.

swagbrdr said:

Bingo. Your basically taking a sample. Any higher sample will be more statistically viable, obviously.

Click to expand...

Just to clarify, in statistics a sample is a portion of a population that’s tested. There are formula for calculating the sample size based on a confidence level and margin of error. We believe, but didn’t ask the OP, that he used his entire population to calculate the SDs.
I would interpret your post to mean if I loaded 50 rounds, I’d randomly select 5 to chrono and infer that’s representative of the other 45. Or, choose a higher sample size, and be statistically more significant. I’m in the group that shoots a small number of rounds first as well, but then either right or wrong, adds to that number of rounds over time with additional recorded shots once I find a load I like.

lordpaxman said:

Just to clarify, in statistics a sample is a portion of a population that’s tested. There are formula for calculating the sample size based on a confidence level and margin of error. We believe, but didn’t ask the OP, that he used his entire population to calculate the SDs.
I would interpret your post to mean if I loaded 50 rounds, I’d randomly select 5 to chrono and infer that’s representative of the other 45. Or, choose a higher sample size, and be statistically more significant. I’m in the group that shoots a small number of rounds first as well, but then either right or wrong, adds to that number of rounds over time with additional recorded shots once I find a load I like.

Click to expand...

You’re right. You would need to have a population to sample from.

Maybe a better way to look at it is an experiment. Change one variable at a time until a desired outcome is achieved. Really, that’s more of the realistic way most of us do things in loading.

I found that if you finally get a good load that has an SD you find acceptable, the trick is being able to repeat the numbers and/or the grouping. If so, than it’s a keeper. I find this works for me anyways. Two different loading and shooting sessions with like results. And I still don’t have to load 50 rounds of one batch to achieve my desired outcome.

As another mentioned above, the SD is not the end all be all. It is however, a diagnostic tool that can help you in load development.

colton1337 said:

not sure I have the other important info stored on it anymore...

Click to expand...

Records are key. I'm not one for voluminous notes but I do use a cheap spiral-bound notebook to write down velocities at the range and to trace the holes (works best at midrange, not as effective with ragged holes) at the end of the day. This data helped me to find the answer to slick polygonal 12" twist rifling on a .223 target rifle once: a quicker less forgiving powder, namely a starting load of H4895. Obviously you'll have different parameters with a 22-250 but a quicker powder just might slow it down if that's what you want. Good luck and nice shootin'

spitballer said:

Records are key. I'm not one for voluminous notes but I do use a cheap spiral-bound notebook to write down velocities at the range and to trace the holes (works best at midrange, not as effective with ragged holes) at the end of the day. This data helped me to find the answer to slick polygonal 12" twist rifling on a .223 target rifle once: a quicker less forgiving powder, namely a starting load of H4895. Obviously you'll have different parameters with a 22-250 but a quicker powder just might slow it down if that's what you want. Good luck and nice shootin'

Click to expand...

I do the same. I have a “reloading journal.” My chronograph allows me to print off the results via excel. I print them out and tape them in there.