Range Report Is SD really that telling? Check this out instead

[bohem]

On many reloading forums you see "low SD, Low ES..." posted ALL the time. But let's think about what the SD is really telling us before saying that SD is what we need to consider.

What does Standard Deviation really mean and how useful would it be by itself?

From Wikipedia:

<span style="font-style: italic">In probability theory and statistics, the standard deviation of a statistical population, a data set, or a probability distribution is the square root of its variance. Standard deviation is a widely used measure of the variability or dispersion, being algebraically more tractable though practically less robust than the expected deviation or average absolute deviation.[citation needed]

It shows how much variation there is from the "average" (mean, or expected/budgeted value). A low standard deviation indicates that the data points tend to be very close to the mean, whereas high standard deviation indicates that the data is spread out over a large range of values.</span>

Here's another thing to consider: <span style="font-weight: bold">SD in and of itself is not particularly noteworthy unless you have a good idea of what SD number you need to have a high confidence level</span> and this is where % variance comes in.

SD/MV is far more useful for determining acceptable levels in a load and for comparing various loads for precision level shooting.

To make my argument a little easier to discuss, let's talk about a 308 and a 30-06 shooting the same bullet, the 175 SMK. These are both very common loads and I'll put in what is considered "good" load results with data that I regularly see from my 30-06 and my friend's 308. Assume both loads are top notch and shoot bugholes in the rifles.

308: 175 SMK @ 2650 fps avg, ES = 25fps, SD = 12 fps
30-06: 175 SMK @ 2950 fps avg, ES = 25fps, SD = 12 fps

If you only apply comparison of the SD values these loads seem identical, but look at the % variance calculation instead.

% variance = SD/MV_avg*100

308: 12/2650*100 = 0.453%
30-06: 12/2950*100 = 0.406%

Now, those numbers both look very small, but let's talk about how different they are in relative terms (% difference)

[(0.406-0.453)/0.406]*100 = -11.6%

This means that while the SD comparison says the loads are identical, the 30-06 load actually has 11.6% LESS variance around it's average MV as opposed to the 308.

A more abrupt comparison would be to do a subsonic 308 vs. a full bore 308 with otherwise identical ES and SD numbers.

MV = 1000 fps vs. 2650 fps with SD = 12?

%subsonic = 12/1000*100 = 1.2%
%fullbore = 12/2650*100 = 0.453%

%difference = [(0.453-1.2)/0.453]*100 = 165% LESS variance.

So here's the bottom line:

If you have the same bullet, same load and very similar MV's that you're comparing then saying "my SD and ES are tighter on load A" is acceptable because all the other terms cancel out of the equation.

If you say "This 308 load is tighter in ES and SD than that 300 WM load because it has x fps smaller SD than the 300 WM, even though I'm shooting a 175 SMK and the 300 WM is shooting a 190 or 210 class" you could (small chance) be correct but mostly you're going to be wrong in actual application.

Take the % variance check and the SD check and look at the vertical dispersion at long distance using JBM to verify it for yourself if you feel the need.

The number you should shoot for is % variance of 0.5% or less.

This application of SD/V was something that I've started using based upon the recommendations/discussions with Ply1951Guy.

 

[BryanLitz]

Re: Is SD really that telling? Check this out instead

Very good point.

<div class="ubbcode-block"><div class="ubbcode-header">Quote:</div><div class="ubbcode-body">Take the % variance check and the SD check and look at the vertical dispersion at long distance using JBM to verify it for yourself if you feel the need.</div></div>

Using your example of 2650 and 2950 fps ave velocity for the 175 SMK at 1000 yards:

Using 10 fps SD for both loads and 95% confidence limits (+/- 20 fps from mean) yields:
15.4" of vertical dispersion for the .308
10.6" of vertical dispersion for the '06

That's a big difference in actual dispersion with the same SD, which illustrates your point.

Doing the same thing with a fixed 0.5% variation is +/- 13 fps for the .308 and +/- 15 fps for the '06. Corresponding vertical dispersions are:
10" for the .308
7.9" for the '06

That's a closer comparison than using the same SD for both, which again illustrates your point.

We agree that the % variation is a closer apples-to-apples comparison of vertical dispersion. I approach the analysis differently than to say: "shoot for a fixed SD or % variation".

My recommended approach is to minimize variation as much as possible, and do the analysis to see how much vertical dispersion this corresponds to for your specific shooting application. For example, of someones load has 1% velocity variation and he's hunting elk out to 600 yards max, that 1% MV variation is going to keep him in the kill zone. So for that application, 1% is good enough and there's no need to get any better.

In my application of FULLBORE shooting, our X-ring is 10" at 1000 yards. Shooting 155's at ~3000 fps, staying within the 1 MOA X-ring happens to require an SD of 10 fps or less (95% confidence) so that's what we typically try to achieve. F-class shooters have an X-ring 1/2 the size so their desired variation should be tighter.

My point is that the degree of MV variation that an individual shooter should find acceptable depends directly on the size of the target they need to hit. For some cartridge/bullet/target combinations, the acceptable limit might be 0.5%, while it might be 0.3% or 1.0% for others.

I would encourage each shooter to do the analysis for their own application to decide what their required threshold is and work on load development until it's achieved (and likewise don't feel obligated to keep working to minimize it once the sensible threshold is met).

Thanks for shining a light on this subject that's often not fully thought out.

-Bryan

 

[bohem]

Re: Is SD really that telling? Check this out instead

Thanks! Nice examples as well, I appreciate it.

ETA: Your approach to what level of variance is acceptable makes a lot of sense, I like it. I've been checking my repeatability window against the targets that I engage on a regular basis and have found that 0.5% or a little better generally gets me what I need but I do like how you approached it, that makes the most sense. Thanks.

 

[wolskyr]

Re: Is SD really that telling? Check this out instead

Tag for reference

 

[azimutha]

Re: Is SD really that telling? Check this out instead

I hadn't considered viewing SD before as a percentage of MV. Thanks for the insight. Aside from a percentage that I can live with, some other factors I've considered have actually led me to choose a load that has more SD than if done with another powder. In a narrow range of temperatures (45F-70F), powder "A" has better SD #s but powder "B" I'm using is less temperature sensitive. I took powder "A" & "B" loads out when it was 15F and 90F to 300 yards. "B" varied less in both cases on the target. It also burns cleaner. And, other than segregating brass, two different rifles shoot the exact same load quite well...makes handloading much less of a chore.

 

[Iggy]

Re: Is SD really that telling? Check this out instead

Thanks! That's great info.

 

[sharac]

Re: Is SD really that telling? Check this out instead

Statistics basically. And also consider using larger samples (when i do load development i usually shoot 5 shot groups - way too small) as small samples are not really representative of entire population. I think best compromise between $$$ and statistical accuracy is to do initial testing on 5 shot (see the OCW/Ladder results, do the stat. math) and then after picking up one or two possible final loads do a match simulation or the way you are going to use that load and get the numbers and redo the math.

Also beneficial is to make a graph of your shot speeds from match simulation (you shoot all rounds as in a match with sighters and everything and record speed and plot them on a chart from 1 - last)as picture says 1000 words. As you do that several times (different conditions) you can make yourself really good dataset to examine and maybe discover how your rifle behaves in different temps. for example in combination with hits plot you can even get a virtual bullet path from 1 - last shot. Still data will be only as good as shooter is as you can't totally eliminate human factor but still with little effort lots can be gained from quality dataset...

 

[Fred_C_Dobbs]

Re: Is SD really that telling? Check this out instead

Occasionally I get a 3-shot group with the exact same MV for all three rounds. IOW, ES and SD both were zero. Despite the fact that all the rifles I reload for are capable of sub-MoA accuracy, none of those 0 ES/0 SD groups has ever been especially small, usually >1 & <2 MoA.

 

[McCrazy]

Re: Is SD really that telling? Check this out instead

Tag. Thanks for a few good nuggets of knowledge gentlemen.

 

[Low Sioux]

Re: Is SD really that telling? Check this out instead

<div class="ubbcode-block"><div class="ubbcode-header">Originally Posted By: Bryan Litz</div><div class="ubbcode-body">Very good point.

For some cartridge/bullet/target combinations, the acceptable limit might be 0.5%, while it might be 0.3% or 1.0% for others.

</div></div>

The question then becomes how much time and effort in reloading and load testing does it take to get from 1.0% to .5% and then to go from .5% to .3%?

 

[bohem]

Re: Is SD really that telling? Check this out ins

Fred- A 3 shot group is useful for getting in the ball park but it's certainly not useful for measuring velocity and saying "this is a good combo to minimize variance". The sample size is much too small.


LS- If you're efficient about the reloading process and using quality components the time to get down to 1% or 1/2% is cut significantly.

Some would argue that there are certain chamberings which will rarely see a 0.3% variance no matter what you load it with. The 308 and most other common calibers that we shoot are generally pretty nice to load for. The 308 in particular is so popular and so well developed that many loads can be tailored to fit the bill.

The best answer I can think of directly to your question is that it's going to come down to your capabilities to develop a load (load and test skills). A fresh out of the gates reloader doing this for the first time might stumble on a fantastic combination through loading manuals or internet advice but they're going to be learning to load just as much as they are going to develop the load.

Someone with years of match ammo tuning and competition development will be able to tune those final tenths down because they know where to look for consistency right away.

 

[Greg Langelius *]

Re: Is SD really that telling? Check this out ins

It's a big world, we shoot small bullets, the error factors in our recorded numbers can be bigger than the numbers we're looking for, and our calculations make assumptions about conditions that are almost never consistent.

Simply put, I doubt the numbers we get from our chronographs can be trusted at a level of precision that makes our calculations geninely meaningful.

Before we calculate anything, let's examine the chronograph error factor. My guess is that somewhere in the accompanying literature is the caveat stating that numbers reported are accurate to +/-1% of the number displayed.

This means, in English, that a velocity reported as 3000fps is actually a number somewhere in between 2970fps and 3030fps.

Now run your dispersion calculation for an average SD of 60fps. Then run it for the reported SD added to that 60fps.

Even if the average reporting error is only half of that 1%, any number you input is only reliable to about 30fps, so just how in heck can numbers smaller than that be even considered as anything but purest fantasy?

When I want to know what my load is doing, I'm looking at the target. The chrono is just another piece of lyin' bastich instrumentation, put on this earth to distract us from the target, which is where the real story unfolds.

IMHO...

Greg

 

[bohem]

Re: Is SD really that telling? Check this out ins

Greg, those are some good points to consider. Here's some more information on them from my investigation.

The reported precision interval on my chrono is +/- 0.005fps/fps = 0.5% on a 3 sigma interval.

1% total variance on 3000fps is now +/- 15fps from mean on a 99.7% confidence interval as opposed to the 2 sigma (95%) interval that Bryan and I discussed for the SD related dispersion. This is also not directly related to SD, it affects it but it is part of the error analysis involved with the ES instead. The SD effects fall out of that but it is not a 1:1 correlation to SD changes.

 

[vman]

Re: Is SD really that telling? Check this out ins

Greg you chrono basher you... hahaha

Top read fellas, and i have to agree with bohem and greg on verification of data via real life testing.

I never truly understood the meaning of consistency until I joined these forums!

 

[MitchAlsup]

Re: Is SD really that telling? Check this out instead

<div class="ubbcode-block"><div class="ubbcode-header">Originally Posted By: bohem</div><div class="ubbcode-body">%difference = [(0.453-1.2)/0.453]*100 = 165% LESS variance.</div></div>

The way I was taught:: 100% less variance is ZERO. You start with a variance, and have 100% of it. Now you take away 100% of that from it and what do you have left? ZERO. There is no such thing as 165% less variance, so you set up that equation wrong.

{Note: the rest of the subject mater is spot on.}

 

[bohem]

Re: Is SD really that telling? Check this out instead

<div class="ubbcode-block"><div class="ubbcode-header">Originally Posted By: MitchAlsup</div><div class="ubbcode-body"><div class="ubbcode-block"><div class="ubbcode-header">Originally Posted By: bohem</div><div class="ubbcode-body">%difference = [(0.453-1.2)/0.453]*100 = 165% LESS variance.</div></div>

The way I was taught:: 100% less variance is ZERO. You start with a variance, and have 100% of it. Now you take away 100% of that from it and what do you have left? ZERO. There is no such thing as 165% less variance, so you set up that equation wrong.

{Note: the rest of the subject mater is spot on.} </div></div>



I'm referencing the % difference of % variance. The % variance that I calculated is then correlated to the other % variance. As I understand it you cannot have more than 100% less variance, however, you can have more than 100% difference in the % variance; which is what I actually calculated. Calling it less variance should have been termed more appropriately as "165% smaller % variance".

It's a normalizing comparison with the % difference calculation. The normalized (unit value) is the supersonic ammo and the subsonic ammo is compared to it.

 

[sharac]

Re: Is SD really that telling? Check this out instead

I think the point of this post was to illustrate that direct comparison of SDs between loads is erroneous and that some people forget to take into account values of their measurements (speeds).

 

[txrpls]

Re: Is SD really that telling? Check this out instead

Great read. I am a land surveyor and I use a least squares adjustment package for our survey measurements. Comparing two loads based on standard deviation without comparison to a standard is actually kind of like comparing apples to oranges. All of the software I have used ask for weighing of the measurement type and also weights according to the length and type of measurement. a qualified GPS vector has more weight than a total station and a total station measurement to a fixed target has more weight than a measurement to a target on a hand held prism pole. All of the measurement could have the same SD, but are not weighted the same. I would think (and I am not a statistician) that for SD to me a valid comparison there would have to be a standard set for a base line measurement.

 

[Low Sioux]

Re: Is SD really that telling? Check this out instead

Good Topic good discussion. I think this whole discussion illustrates why most most professional reloaders, I mean bench rest shooters,focus on ES rather than on SD. The simple idea of minimizing the velocity spread gets you to where you want to be; namely minimal vertical spread without having 3 college classes in statistics

While this is certainly a worthy goal, and absolute must for the elite competitor, I have to ask myself if all the fuss is worth it for the average "redneck rock killer" like myself. As Greg pointed out, is your $119 chrono accurate enough to be meaningful? if the error is always 20fps fast or 20fps slow maybe, but the only way to really know is to match you chrono results with what you see at the target.

Then comes the wind.
I for one have never shot on the groomed lawns of a camp perry or similar ranges with wind flags and at least a somewhat predictable wind and terrain features. Out here we shoot over LOTS of uneven terrain features that are giving you vertical wind components. I cant measure those. With experience I can guess what they are going to be with a fair amount of success, but I cant measure them and therefore even though they are relevant I cant make them statistically relevant.

I guess my point here is I wont spend a lot of time trying to fix something that may not be broken. If I am consistently missing .25 mil rocks vertically (one shot over one shot under) more often than from wind My load may need some work, but then how can I accurately measure my success other that at the target? And if takes turning necks and soft seating bullets to get there i am just not willing to do it.

For most of us non competitive shooters it may make more sense with a day to burn to load a bunch "good" ammo in the morning and shoot it in the afternoon, than to spend the entire day sorting, weighing, and turning necks in the hopes of "great" ammo. If you dont shoot enough to learn to dope the wind "great ammo" misses just as often as "good" ammo. Go Shoot

 

[sharac]

Re: Is SD really that telling? Check this out instead

Magic word: Consitency.... whatever you do while shooting (EVEN if you do everything wrong - don't take that literally ) you'll get good data if you do it consistently...

So absolute numbers (of your load) are totally irrelevant as long as you always use same methods of measuring (same chrony, same shooting, same etc...) yeah there is an built-in error but if the same error is present in every load/statistical sample end result is the same. Just don't consider statistics as an absolute science cause it aint

...

To illustrate: In statistics you can have totally different time series (absolute measure) and they can even be totally wrong (e.g. a research with error in data entry) however if this error is consistent (e.g. error made is always the same) you get the same descriptive data (e.g. trend).

 

[Greg Langelius *]

Re: Is SD really that telling? Check this out instead

This is precisely why I flatly refuse to get insanely anal about my handloading techniques. I would be proud as Punch to get 1/2MOA out of my system at 200yd, but I recognize that I'm way far from alone in not being able to even achieve 1MOA in the real world (40 shots into 2" at 200yd).

Real world accuracy is about puttin' 'em all where you want 'em. If I were hunting, every danged thing I was tryin' to kill would be dead.

When I designed the FV200 target I did it so it would be an achieveable challenge, but only just barely. It's primary goal is to permit our shooters to demonstrate to themselves just exactly how good they are, or aren't. Real world, not shooting groups from a bench.

For my purposes, 1MOA (inside the 2" 5-ring) is highly satisfactory, 1/2MOA (inside the 1" V-ring) is outstanding, and anything smaller is essentially the stuff of fiction. I don't try for the fiction. I sometimes get 2MOA (everything inside the 4" 4-ring), but mostly it's not even that good. That's <span style="font-style: italic">my</span> real world and it keeps on bringin' me back to try again. I make no excuses, I'm doing my best, and I'm not going to quit trying to do better. It <span style="font-style: italic">will</span> happen.

Greg

 

[sobrbiker883]

Re: Is SD really that telling? Check this out ins

<div class="ubbcode-block"><div class="ubbcode-header">Originally Posted By: Greg Langelius *</div><div class="ubbcode-body"><span style="font-weight: bold">When I want to know what my load is doing, I'm looking at the target. The chrono is just another piece of lyin' bastich instrumentation, put on this earth to distract us from the target, which is where the real story unfolds.

IMHO...

Greg</span> </div></div>


Your humble opinion is spot on, as usual (at least IMHO). You sir, have a firm grasp of functional practicality and its application to putting rounds on a target. Thank you for your continuing contributions!