Required reading, RSD vs extreme spread.

Ammunition Discussion.

Moderators: mpallett, bakerjw

Post Reply
User avatar
silencertalk
Site Admin
Posts: 33978
Joined: Mon Apr 25, 2005 8:31 am
Location: USA

Required reading, RSD vs extreme spread.

Post by silencertalk »

http://www.shootersjournal.com/Features/WHICHONE.pdf

It shows that using radial standard deviation you can determine which load is best with 40% greater accuracy than using extreme spread EVEN if 25 shots are fired!

It also shows that with 3 shot groups, you can never be more than 47% certain of which load is best. With 10 shot groups, you can be 75% certain, and with 30 shot groups, you can be 90% certain.

http://fuzzylimey.net/coachtalk/skillanal.html
apochachuva
Silent But Deadly
Posts: 2338
Joined: Mon Jul 23, 2007 9:38 pm
Location: Houston, Texas

Re: Required reading, RSD vs extreme spread.

Post by apochachuva »

rsilvers wrote:http://www.shootersjournal.com/Features/WHICHONE.pdf

It shows that using radial standard deviation you can determine which load is best with 40% greater accuracy than using extreme spread EVEN if 25 shots are fired!

It also shows that with 3 shot groups, you can never be more than 47% certain of which load is best. With 10 shot groups, you can be 75% certain, and with 30 shot groups, you can be 90% certain.

http://fuzzylimey.net/coachtalk/skillanal.html
Yay, confidence intervals.
"You'll put yer eye out with that thing"

"How a politician stands on the Second Amendment tells you how he or she views you as an individual; as a trustworthy and productive citizen, or as part of an unruly crowd that needs to be lorded over, controlled, supervised, and taken care of." - Dr. Suzanna Gratia Hupp
User avatar
silencertalk
Site Admin
Posts: 33978
Joined: Mon Apr 25, 2005 8:31 am
Location: USA

Post by silencertalk »

5 shot groups, which seem the most common, are about 57% reliable. Basically everyone needs to switch to 10 shot groups if they are casual and don't care that much, or 30 shot groups if they want to be pretty reliable. And also switch to mean-radius.
apochachuva
Silent But Deadly
Posts: 2338
Joined: Mon Jul 23, 2007 9:38 pm
Location: Houston, Texas

Post by apochachuva »

rsilvers wrote:5 shot groups, which seem the most common, are about 57% reliable. Basically everyone needs to switch to 10 shot groups if they are casual and don't care that much, or 30 shot groups if they want to be pretty reliable. And also switch to mean-radius.
I'm going away from benchrest shooting and am going to try more standing shooting.
I already tested my handloads by firing 20-30 shot strings, but I never really had any reason other than that's the way I wanted to do it.
"You'll put yer eye out with that thing"

"How a politician stands on the Second Amendment tells you how he or she views you as an individual; as a trustworthy and productive citizen, or as part of an unruly crowd that needs to be lorded over, controlled, supervised, and taken care of." - Dr. Suzanna Gratia Hupp
User avatar
silencertalk
Site Admin
Posts: 33978
Joined: Mon Apr 25, 2005 8:31 am
Location: USA

Post by silencertalk »

Ok. Well with your 20-30 shots, you can classify results 40% more accurately if you do it with the RSD method. RSI Shooting Lab software can do this. It is $100. There is a free demo. It can also do internal and external ballistics. I already own QuickLoad for internal ballistics and love it.
User avatar
silencertalk
Site Admin
Posts: 33978
Joined: Mon Apr 25, 2005 8:31 am
Location: USA

Post by silencertalk »

User avatar
silencertalk
Site Admin
Posts: 33978
Joined: Mon Apr 25, 2005 8:31 am
Location: USA

Post by silencertalk »

johndoe3
Silent But Deadly
Posts: 2710
Joined: Tue Feb 27, 2007 3:02 am
Location: N. Colorado

Post by johndoe3 »

Unless one is shooting from a complete mechanical rest where the rifle is clamped down to give the same shot every time, then shooter fatigue enters the issue (for a 30 shot group) and causes larger size groups than would normally be experienced.

Personally, I think 5-6 groups of 5 shots each gives a more "real life" possibility and avoids shooter fatigue (sloppiness in the later shots when shooting 20-30 shots). If you shoot 5-6 groups of 5 shots each, the average is not your capability, it is the largest two groups (which you could average). With a sample of 6 groups of 5 shots, you have a good understanding of the rifle and your own capability. That way you know that your rifle will shoot this group size or better unless you muff a shot or have ammo that is way outside the normal SD.

I think a statistical sample size of 6 (groups) is superior to a sample size of 1 group (of 30 shots), solely because of the shooter fatigue issue and perhaps also barrel heating.
User avatar
silencertalk
Site Admin
Posts: 33978
Joined: Mon Apr 25, 2005 8:31 am
Location: USA

Post by silencertalk »

You can combine targets. I don't care how much time it takes to get the 30 shots I feel are needed. 6 groups of 5 shots is ok.

So who is up for writing a web app which can do all these statistics? I never learned how to code for web apps.
User avatar
chrismartin
Silencertalk Goon Squad
Posts: 4226
Joined: Sun Mar 12, 2006 7:18 pm
Location: Tidewater, VA

Post by chrismartin »

rsilvers wrote:You can combine targets. I don't care how much time it takes to get the 30 shots I feel are needed. 6 groups of 5 shots is ok.

So who is up for writing a web app which can do all these statistics? I never learned how to code for web apps.
I learned to write web apps, but never to do math :)

Seriously, I can write the app if someone can help with the math.

Plus, I can make it iPhone/IpodTouch compatible too, though I'm not an iPhone developer, so I can't integrate with the camera.
User avatar
silencertalk
Site Admin
Posts: 33978
Joined: Mon Apr 25, 2005 8:31 am
Location: USA

Post by silencertalk »

The math is the easy part and I can show you how if you collect the x,y position of each shot.

To start, download and try RSI Shooting Lab and see how it works. Back when I was in my prime I could write the target part of it in one programming session.
apochachuva
Silent But Deadly
Posts: 2338
Joined: Mon Jul 23, 2007 9:38 pm
Location: Houston, Texas

Post by apochachuva »

chrismartin wrote:
rsilvers wrote:You can combine targets. I don't care how much time it takes to get the 30 shots I feel are needed. 6 groups of 5 shots is ok.

So who is up for writing a web app which can do all these statistics? I never learned how to code for web apps.
I learned to write web apps, but never to do math :)

Seriously, I can write the app if someone can help with the math.

Plus, I can make it iPhone/IpodTouch compatible too, though I'm not an iPhone developer, so I can't integrate with the camera.
The math is pretty easy, The hardest part being calculating variance. You could manually enter the coordinates if you used targets with a grid.
in this case I think you can simply use the following equations
group x mean = sum of all x coordinates /# of total shots
group y mean = sum of all y coordinates /# of total shots
RSD=sqrt( Sum for all x's of (|x coordinate-group mean X coordinate|)^2/(# of shots) + Sum for all y's of (|y coordinate-group mean y coordinate|)^2/(# of shots)

You could do this with a php form pretty quick to take care of the math, but even just automating it could be a handy little app to have on your phone.

For the camera/scanner approach you need to enter a scale of some sort.

ETA:We could simply start using graph paper for targets to facilitate manual data entry.
"You'll put yer eye out with that thing"

"How a politician stands on the Second Amendment tells you how he or she views you as an individual; as a trustworthy and productive citizen, or as part of an unruly crowd that needs to be lorded over, controlled, supervised, and taken care of." - Dr. Suzanna Gratia Hupp
User avatar
silencertalk
Site Admin
Posts: 33978
Joined: Mon Apr 25, 2005 8:31 am
Location: USA

Post by silencertalk »

You could calibrate your screen by holding a CD-ROM up to the monitor and then dragging a circle to the same size. It could then output your horz and vert dpi. You could then just enter those values in next time or it could save them in a cookie.

If it was an iPhone app, it could ask you to place a quarter on the target and then photo it. You could then touch the screen where the quarter is, and it would calibrate. And then touch each shot hole. You would need to be able to pan and zoom.

With RSI Shooting lab, you hold the target on your monitor and then just click on the bullet holes. It works well. Alternatively you could scan the target but I think that is too much work and I would not do it that way.
apochachuva
Silent But Deadly
Posts: 2338
Joined: Mon Jul 23, 2007 9:38 pm
Location: Houston, Texas

Post by apochachuva »

rsilvers wrote:You could calibrate your screen by holding a CD-ROM up to the monitor and then dragging a circle to the same size. It could then output your horz and vert dpi. You could then just enter those values in next time or it could save them in a cookie.

If it was an iPhone app, it could ask you to place a quarter on the target and then photo it. You could then touch the screen where the quarter is, and it would calibrate. And then touch each shot hole. You would need to be able to pan and zoom.

With RSI Shooting lab, you hold the target on your monitor and then just click on the bullet holes. It works well. Alternatively you could scan the target but I think that is too much work and I would not do it that way.
I like the iPhone app idea best of these. I'm going to get the SDK and mess around with it, see if I can find out how difficult it would be.
ETA: The SDK is Mac only, so I'll have to find a Mac to work on.
"You'll put yer eye out with that thing"

"How a politician stands on the Second Amendment tells you how he or she views you as an individual; as a trustworthy and productive citizen, or as part of an unruly crowd that needs to be lorded over, controlled, supervised, and taken care of." - Dr. Suzanna Gratia Hupp
apochachuva
Silent But Deadly
Posts: 2338
Joined: Mon Jul 23, 2007 9:38 pm
Location: Houston, Texas

Post by apochachuva »

Update, so for this semester I have to do some programming on iPhones anyway, so I'll be writing this app if I get enough spare time. I don't know if I'll be able to publish it through iTunes or not though.
"You'll put yer eye out with that thing"

"How a politician stands on the Second Amendment tells you how he or she views you as an individual; as a trustworthy and productive citizen, or as part of an unruly crowd that needs to be lorded over, controlled, supervised, and taken care of." - Dr. Suzanna Gratia Hupp
User avatar
silencertalk
Site Admin
Posts: 33978
Joined: Mon Apr 25, 2005 8:31 am
Location: USA

Post by silencertalk »

Thanks. I was banned on Tactical Forums for this post. The moderator did not understand it and lunged for me and I wrote back and he is not used to being wrong so could not take it.
apochachuva
Silent But Deadly
Posts: 2338
Joined: Mon Jul 23, 2007 9:38 pm
Location: Houston, Texas

Post by apochachuva »

rsilvers wrote:Thanks. I was banned on Tactical Forums for this post. The moderator did not understand it and lunged for me and I wrote back and he is not used to being wrong so could not take it.
I guess he didn't take statistics.
"You'll put yer eye out with that thing"

"How a politician stands on the Second Amendment tells you how he or she views you as an individual; as a trustworthy and productive citizen, or as part of an unruly crowd that needs to be lorded over, controlled, supervised, and taken care of." - Dr. Suzanna Gratia Hupp
User avatar
silencertalk
Site Admin
Posts: 33978
Joined: Mon Apr 25, 2005 8:31 am
Location: USA

Re: Required reading, RSD vs extreme spread.

Post by silencertalk »

Now I am thinking just normal radial standard deviation times two for no less than 10, but ideally 30 shots to get a diameter. The values will be small enough to not scare away people/makers from quoting them (it will predict a typical 4 shot group). And you can multiply them by 3 to get the universe of shots if you wish.

The problem is, gun makers (and shooters) will just fire lots of 3 shot groups and quote the BEST one rather than the aggregate of them all. Really even a crappy rifle will shoot a 0.25 MOA 3 shot group once in a while.
User avatar
silencertalk
Site Admin
Posts: 33978
Joined: Mon Apr 25, 2005 8:31 am
Location: USA

Re: Required reading, RSD vs extreme spread.

Post by silencertalk »

One final idea - call it RSD2 rather than DSD. That would still look like an expected diameter value but would really show the radial distribution for the second standard deviation (95% of the shots). That satisfies the people who don't understand the number and they just see what looks like a typical extreme spread value but also satisfies those who know it to be a radius for 95% of the universe of shots rather than a diameter.
apochachuva
Silent But Deadly
Posts: 2338
Joined: Mon Jul 23, 2007 9:38 pm
Location: Houston, Texas

Re: Required reading, RSD vs extreme spread.

Post by apochachuva »

Unfortunately I never had much time to do anything other than the required coursework. I'd still like to see this app written.
"You'll put yer eye out with that thing"

"How a politician stands on the Second Amendment tells you how he or she views you as an individual; as a trustworthy and productive citizen, or as part of an unruly crowd that needs to be lorded over, controlled, supervised, and taken care of." - Dr. Suzanna Gratia Hupp
User avatar
JohnInNH
Elite Member
Posts: 3313
Joined: Thu Jul 19, 2007 10:49 pm
Location: SW NH

Re:

Post by JohnInNH »

silencertalk wrote:Ok. Well with your 20-30 shots, you can classify results 40% more accurately if you do it with the RSD method. RSI Shooting Lab software can do this. It is $100. There is a free demo. It can also do internal and external ballistics. I already own QuickLoad for internal ballistics and love it.

I have both and find them each good for different things.
Long distance, the next best thing to being there!
Post Reply