Standard Deviation vs Extreme Spread

Discussions of powders, bullets and loading information.

Moderators: Kirk, Lucinda

mdeland
Posts: 11553
Joined: Mon Nov 20, 2006 1:47 pm

Re: Standard Deviation vs Extreme Spread

Post by mdeland » Wed Dec 25, 2019 4:40 pm

Good post Bruce , your ideas were easy to follow and you articulate very well. I'm still not convinced that standard deviation is as relevant to a loads "cone of dispersion" as is the extreme spread. True it's only two samples in a groups execution but those two very often are the wide shots in the group or a miss at 1K.

bruce m
Posts: 3351
Joined: Tue Aug 04, 2009 5:25 am
Location: australia

Re: Standard Deviation vs Extreme Spread

Post by bruce m » Wed Dec 25, 2019 4:51 pm

mike,
if anyone understands statisics and s.d. on this forum, it is brent.
bruce.
ventum est amicus meus

Aviator
Posts: 285
Joined: Mon Sep 03, 2018 8:06 pm

Re: Standard Deviation vs Extreme Spread

Post by Aviator » Wed Dec 25, 2019 10:01 pm

Well, I may be the oddball here, but I am in the habit of looking at both extreme spread and standard deviation of the velocities in a group.
They are both indicators of consistency, and are both part of the story in analyzing how well your load is working.
They are not the same, but in general track together. You will never get really low standard deviation without also having low extreme spread, or really low extreme spread without also having low standard deviation. I do not have any reason to ignore one or the other, they are both good information. Additionally, I pay attention to the raw data of individual shots, to see if a large extreme spread or standard deviation is driven by just one aberrant shot in a group.

One important thing to remember is that if you are trying to draw conclusions from a data sample, a larger sample size is better. The larger the sample size, the higher confidence you can have that your sample truly approximates the full range of velocities that your load will produce. For instance, if something about your rifle, brass, primers, powder, powder measurement, bullet, overall length, etcetera, gives you an aberration in velocity about 1 in 10 shots, a 3 or 5 shot group is not going to be adequate for drawing conclusions....

There is rarely a concrete answer to anything in Statistics.....
But better information can result in better assumptions....

(I don't have access to the article that was referenced.)

User avatar
desert deuce
Posts: 3089
Joined: Mon Apr 14, 2008 10:51 pm
Location: Rio Rico, Arizona

Re: Standard Deviation vs Extreme Spread

Post by desert deuce » Thu Dec 26, 2019 12:49 am

There goes Aviator, thinking again. Here I am trying to help folks improve their ammunition and he wants to take us to Engineering Class at the University. Think he should change his forum handle to Professor. His brain housing group is likely why he has progressed so rapidly in this sport.

Anyway, for you out there with a chronograph that expend all of one lot of powder and crack a new case of powder you should know the volume you used for your favorite long range load and how much it weighed.

Load up 5 rounds just like you would with the old lot by volume with the new lot. The weight will likely be different from the old lot. For instance, if the new load by volume weighs 78.5 grains that is your baseline, the next five will be 80.0 grains and the next five 81.5 grains.

That is Laddering up in 1.5 grain increments. This will give you an idea how much velocity is gained per grain of powder that load. The only thing you are changing is the powder charge and amount of compression. Everything else the same.

At this point you should have 15 loaded rounds. Shoot them in their individual five shot groups over the chronograph. Of course the chronograph will read out and most will record each shot in sequence.

So the chronograph says the 80.0 grain load reads, 1325, 1325, 1326, 1327, 1324 That is an ES of 3. This a preliminary indication this load is worth tinkering with. The other two loads were in the category of 14 ES.

At this point I load ten of the 80.0 grain loads and back to the chronograph, 1319, 1319, 1319,1320, 1320,1323, 1322, 1325, 1325,1325 which gives us an ES of 6. An ES in single digits is good. Works best if rifle and ammunition are not in direct sun light. Radiant and/or ambient heating can affect velocity.

Now this looks pretty consistent so load up 30 like these ones and head for the range and shoot them on paper, I first shoot ten at 800 and then ten at 1,000. If the load is shooting to call I don't think I can ask for better results so we take that load to the next match and find out how it does in competition.

The idea here is how to cut down on driving time, gas, alloy, powder and primer expenditures and get to a useful load when the range is 2-3 hours away. With this much work done before you get to the range the shooting at the range will be much more efficient.
Sometimes you get the chicken, and sometimes you get the feathers!

Aviator
Posts: 285
Joined: Mon Sep 03, 2018 8:06 pm

Re: Standard Deviation vs Extreme Spread

Post by Aviator » Thu Dec 26, 2019 7:55 am

HA, you sly 'ol desert deuce!
It doesn't go that smoothly for me, so maybe you're on to something.... :D
Studying Statistics always did give me a headache.... :lol:

User avatar
desert deuce
Posts: 3089
Joined: Mon Apr 14, 2008 10:51 pm
Location: Rio Rico, Arizona

Re: Standard Deviation vs Extreme Spread

Post by desert deuce » Thu Dec 26, 2019 10:13 am

Studying Statistics always did give me a headache.
Boy, can I relate to that headache business in relation to statistics, more like a migraine. Remembering first day of a 'required' statistics class my junior year at the University of Florida. Walked into a large auditorium and looked down at a distant stage with a podium at the center. The auditorium was packed. About ten minutes after class was scheduled to begin two men walked to the podium and addressed the assembly. One held up the text for the class and said, " Welcome to STAT 365, I am Doctor so & so, I wrote this text book you will use for this quarter in STAT 365. One week from today you will have a test on the material covered and after that test 300 of you will be gone, one way or another. This is my teaching assistant Joe Schmoe, he will conduct this class and it is his responsibility to ensure that 300 of you are gone in a week." The professor turned and walked off the stage and we never saw him again that quarter. Nice personal touch don't you think?

Joe Schmoe stepped up to the microphone and said, "open your texts to page so & so," and started right in as the rustling of papers and books reverberated through the auditorium and a lot of people got up and left. I think over 300 dropped the course before the dreaded first test and even more evaporated after that first test. By the time drop day rolled around it seemed about half the class was gone.

Joe Schmoe was actually a pretty good teacher even if he seemed not to have a soul when at the podium. (I suspected he had committed the text book to memory.) During office hours and help sessions he approached human characteristics. Occasionally, especially at office hours, Joe would leak some real pearls of wisdom about statistics. My favorite was junk in junk out, another memorable one was liars figure and figures lie. If you really wanted Joe to wax doctoral ask about "Sampling Error" or "Sampling Bias." It was at office hours it was discovered that Joe had submitted his doctoral thesis and was waiting out formalities before receiving his Doctoral Degree in Statistical Analysis. Essentially, Sampling Error was just that an error in collecting or defining the sample, Sampling Bias on the other hand was collating information to produce a desired result and finally at the end Joe talked about searching for useful information that could be found, if you knew how to find it, in either. Useful information provided a pathway to correct decision making. The word inference came up a lot and occasionally seemed like an excuse for a result.

So in leaning toward Extreme Spread, and remembering Joe, I naturally grasp at the useful information to take me to the desired result of consistent ammunition. Same as statistical analysis, if the procedure/process proves valid, use it. I think of it as a practical approach to useful statistics with ES providing me the most useful information during a process of load development. Perhaps not very scholarly but for me it works. What you do with information is entirely up to you. To me theory is not near as interesting as achievement.
Sometimes you get the chicken, and sometimes you get the feathers!

mdeland
Posts: 11553
Joined: Mon Nov 20, 2006 1:47 pm

Re: Standard Deviation vs Extreme Spread

Post by mdeland » Thu Dec 26, 2019 2:23 pm

Dang, and to think I came to the same conclusion without a degree! :lol: It takes real guts and some brains to get through statistics as some of you folks have. I would have been the first one of the 300 out the door I'm afraid.

User avatar
desert deuce
Posts: 3089
Joined: Mon Apr 14, 2008 10:51 pm
Location: Rio Rico, Arizona

Re: Standard Deviation vs Extreme Spread

Post by desert deuce » Thu Dec 26, 2019 6:13 pm

Trust me Mike, what I learned in (sadistics as I called it) was not worth the trouble or the headaches in particular.

The Hell of it was you had to pay to be tortured.
Sometimes you get the chicken, and sometimes you get the feathers!

Aviator
Posts: 285
Joined: Mon Sep 03, 2018 8:06 pm

Re: Standard Deviation vs Extreme Spread

Post by Aviator » Thu Dec 26, 2019 6:13 pm

desert deuce wrote:
Thu Dec 26, 2019 12:49 am
.... Think he should change his forum handle to Professor.....

Naw, Professor Farringer is my father.

I am just a BS. 8)




(Bachelor Science- Physics)

BFD
Posts: 2789
Joined: Mon Jan 21, 2008 7:36 pm

Re: Standard Deviation vs Extreme Spread

Post by BFD » Thu Dec 26, 2019 6:20 pm

Yup, statistics is a complete waste of time. That's why they invented the field, dedicate billions of dollars to it in industry, education, and the military just to torture people with it. Indeed. Completely lunacy.

Of course, just maybe there is something to all of that, after all.

Nah... Can't be.

;)

mdeland
Posts: 11553
Joined: Mon Nov 20, 2006 1:47 pm

Re: Standard Deviation vs Extreme Spread

Post by mdeland » Thu Dec 26, 2019 7:05 pm

I think it's valuable but I'm not educated enough to put the info to practical use so have to go to plan "B" (ES) to get a better handle on what I'm trying to figure out.
My son , a mining Geologist, had to have both Calculus and Statistics to graduate from the U of AK. He said statistics was the harder of the two.

BFD
Posts: 2789
Joined: Mon Jan 21, 2008 7:36 pm

Re: Standard Deviation vs Extreme Spread

Post by BFD » Thu Dec 26, 2019 7:11 pm

A mining geologist would use things like semivariograms instead of simple standard deviations. Spatial stats or geostats are what keep mining companies from going broke. The people that run such companies are not fools. They use statistical analyses to make decisions. Shooters could learn from that, but they seem generally quite determined not to.

You know what they say about horses. Applies to shooters too.

SchuetzenDave
Posts: 748
Joined: Mon Jan 07, 2013 9:07 am
Location: St. Albert, Alberta

Re: Standard Deviation vs Extreme Spread

Post by SchuetzenDave » Thu Dec 26, 2019 7:56 pm

To fine tune rifles I have found once your SD is less than 1% of your average velocity; reducing the SD may not achieve better accuracy.

Once I obtain an SD of 1% I test 0.1 grain variations by testing multiple 30 shot groups to verify accuracy.

Very frequently a SD of 6 or 7 will result in smaller group sizes than an SD of 3 or 4.

Yes you need to fine tune your rifle to have an SD less than 1% but to fine tune it further you have to compare 30 shot groups sizes for the minor changes to your powder to better improve on accuracy. A reduction in SD at this time may not provide the best accuracy

SchuetzenDave
Posts: 748
Joined: Mon Jan 07, 2013 9:07 am
Location: St. Albert, Alberta

Re: Standard Deviation vs Extreme Spread

Post by SchuetzenDave » Thu Dec 26, 2019 7:56 pm

To fine tune rifles I have found once your SD is less than 1% of your average velocity; reducing the SD may not achieve better accuracy.

Once I obtain an SD of 1% I test 0.1 grain variations by testing multiple 30 shot groups to verify accuracy.

Very frequently a SD of 6 or 7 will result in smaller group sizes than an SD of 3 or 4.

Yes you need to fine tune your rifle to have an SD less than 1% but to fine tune it further you have to compare 30 shot groups sizes for the minor changes to your powder to better improve on accuracy. A reduction in SD at this time may not provide the best accuracy

AzTBH
Posts: 186
Joined: Sun Sep 04, 2005 9:38 am
Location: Peoria, AZ

Re: Standard Deviation vs Extreme Spread

Post by AzTBH » Thu Dec 26, 2019 9:02 pm

I found this Ken Oehler article, from Handloader #138 (March-April 1989), “Standard Deviation … without equations” helpful in explaining standard deviation and its importance, as one of the indicators, in developing a good load.

https://www.riflemagazine.com/magazine/ ... artial.pdf
Ross McCollum
Peoria, AZ

NRA Life Member

Post Reply