The Truth and Nothing but the Truth

JAS-SH

Gunny Sergeant
Full Member
Minuteman
Jun 5, 2020
744
1,288
I'm not only a photographer, I'm an analyst by profession. I gravitate to hard core analysis when I shoot, even keeping a log of every shot and where it went. Anal is the prefix to analyst so sue me! After reading a a very recent and pretty comprehensive article over at PRB on statistical analysis as it concerns rifle shooting I downloaded the free trial for the OnTarget TDS software today. I really recommend this PRB 3 part series. It is chock full of good information. I also recommend that if you're going to read more than one of the three parts nonstop to get a pot of coffee going! There are a lot of quotes included in there from some of the foremost experts on this subject, including Brian Litz, and so I highly recommend the series.

Putting the PRB three part articles and the OnTarget software together was an eye opener. Luckily, I had saved some of my targets from previous shoots so that I was able to use them for testing and analysis. Having saved some of the targets I've shot with my .22 Magnum trainer I was able to put together a picture that matches the articles to a tee. Makes me a believer, at least until I get additional data but that's not likely to contradict the results.

Here are the link to the first PRB article post. You can find the rest on the follow up links:

https://precisionrifleblog.com/2020/11/29/statistics-for-shooters/

Downloaded the software and got to work. I did not cherry pick my groups. Actually, the opposite is true. I can show you some .4 MOA 5 shot groups at 100 yards but I purposely picked the average/worse groups, so I could see what patterns emerged. The software allows you to superimpose multiple groups which is superb. The end result is that you can shoot different groups any way you want, in my case, 2, 3 shot groups per target and aggregate them over a number of targets into one virtual target. Just happens that this is exactly what I had done. I shot 3 shot groups, reloading in between, 2 groups per target. Most of the time......

All shots were made at 100 yards and aimed at 1.5 MOA targets, with a 1/2 MOA center circle and a 1/4 MOA diamond inside of that. Here is what the target looks like (I print my own):

i-2pMfQhh-X2.jpg


OnTarget has a "virtual" target display that does not include the actual target image. Once the software figures out (with your input) where everything is it generates a virtual target/impacts along with a TON of very meaningful data, including mean radius and CEP (Circular Error Probable) numbers. The US Army uses mean radius exclusively in their evaluation of all weapons. We all use extreme spread, which is good, but not really so much for statistical probability of where your next shot will land. This is because extreme spread weighs outliers in the groups much heavier than the rest of the shots. The original is always available if you need to look at it, but for analysis it becomes superfluous This virtual mode is the mode I used for analysis.

So on to the targets. The targets were made shooting 2 - 3-shot groups into 5 separate targets. The number of shots varied and I actually lost a shot somewhere because the numbers of shots do not match multiples of three by two shots.

This is the first virtual target - 2, 3 shot groups with a LOT of important numbers, including mean radius, SD, and CEP. I wont get into that, you'll have to read the articles :)!

i-7VzxTGn-XL.jpg


But look at what happens when I added 4 more 3 shot groups (losing one shot somewhere) for 17 shots:

i-dRt8nMS-XL.jpg


At this point it looked like I had some fliers. Fliers are what most people don't count. In this image it looks like there at 2 fliers (bottom center left). So do I take them out? Look at what happens when I overlay more groups:

i-sF8zRQx-XL.jpg



Now I'm up to 11-3 shot groups. Lost another shot somewhere (again) so it becomes a 31 shot group. The important point here is that the shots that looked like fliers are not anymore. They are grouped so -- no flyers. By the way the light purple circle is not an MOA or inch measurement, it is SD and CEP related. CEP (Circular Error Probable) is an excellent measure of accuracy and precision. I want more!

All the best,

JAS
 
Last edited:
The thing is, if people actually recorded and analyzed all their shots, they would be pretty depressed about how they are shooting. Better to look at the best groups, discount shots out of that group as a 'flier', then be able to claim they have a rifle that 'shoots .5 MOA all day long, if I do my part'. Finding they actually hold 1.2 MOA would be a serious downer.
 
The thing is, if people actually recorded and analyzed all their shots, they would be pretty depressed about how they are shooting. Better to look at the best groups, discount shots out of that group as a 'flier', then be able to claim they have a rifle that 'shoots .5 MOA all day long, if I do my part'. Finding they actually hold 1.2 MOA would be a serious downer.

We've all been guilty of that at one time or another. I'm officially old so I have a new excuse :). That said looking at the CEP (Circular Error Probable) and mean radius numbers after 31 shots I have a 95% probability of hitting a target inside 1 MOA on the next shot Not too bad for an old guy with a non-match rimfire cartridge at 100 yards. I'll take that!
 
Last edited:
  • Like
Reactions: TriggerJerk!
If you want to know how big your group is likely to be based on the mean radius, you can multiply by 2.1 to get the 95% radius (R95). For example, suppose you shoot a 10-shot group at 1000 yards with a mean radius of 3 inches. 2.1 times 3 inches gives you the radius of the circle in which 95% of your shots can be expected to land, so that’s 6.3 inches. Remember that’s a radius, so the total group size would be [the Diameter] around 12.6 inches.
 
I use phone app TargetScan for shot placement recording. It makes testing ammo much easier.

Screenshot_20201223-124334_TargetScan.jpg


At the moment I am recording many types of .223 for my HK.
Screenshot_20201223-124604_TargetScan.jpg

I can see all the groups and data about them quite easily.
Screenshot_20201223-124737_TargetScan.jpg

I like it very much.

Here is my 150m attempt with the Quad near coast, it was quite windy.

Screenshot_20201223-124956_TargetScan.jpg

2 days ago I was there with SK PM and bad winds and got barely 1/3 of shots in the black, some hitting also wrong targets. I burned those targets already. DO NOT TELL ANYONE.
 
I will have to give target scan a try.

I switched to 10 shot groups this summer, it has been humbling, but more valuable as I understand my own capabilities better.
 
Are you sure you're doing this right?

Yes, and I’m basing the 95% data on CEP, NOT mean radius. Your quote on your post below from Brian in part 3 of the article is correct. The 95% number predicting the DIAMETER for THE NEXT GROUP (not the next shot) would exceed 1 MOA. In other words, predicting the next group, and the next shot are not the same thing. And It's all probability not certainties ;).

l checked the article on CEP in Ballistipedia before trusting the number put put by the OnTarget software. What I found is that the math is way over my pay grade!
 
Last edited:
Yes, and I’m basing the 95% data on CEP, NOT mean radius. Your quote on your post below from Brian in part 3 of the article is correct. The 95% number predicting the DIAMETER for THE NEXT GROUP (not the next shot) would exceed 1 MOA. In other words, predicting the next group, and the next shot are not the same thing. And It's all probability not certainties ;).

l checked the article on CEP in Ballistipedia before trusting the number put put by the OnTarget software. What I found is that the math is way over my pay grade!

CEP is a radius not a diamter.

Nobody quotes accuracy with radius, since common parlance of "1 MOA" is a measure of extreme spread of a group. If we use this to form the measurement of a circle, the circle we create about the maximal dispersion is the circle's "maximum chord"

A circle's maximum chord is its diameter, not its radius.

diameter
A chord that passes through the center of a circle is called a diameter and is the longest chord.

It goes without saying that the radius is not equal to the diameter, since its off by a factor of 2x.

This is important geomtry 101, its not being pedantic.
 
There is one takeaway from this exercise that I can be happy about. If you look at the data and the virtual images, you'll see two crosshairs - a purple one and a green one. The purple is point of aim, the green is center of the group(s), or X,Y offsets from POA.

I 'm happy in that as I added more and more shots the offset (center of the group(s)) shifted very little. That's important to me. In this case the X,Y offsets never exceeded .1 mil from the aim point, and the Y offset (elevation) was very, very consistent. So my aim was pretty consistent even if my other fundamentals were "not so much".......
 
CEP is a radius not a diamter.

Nobody quotes accuracy with radius, since common parlance of "1 MOA" is a measure of extreme spread of a group. If we use this to form the measurement of a circle, the circle we create about the maximal dispersion is the circle's "maximum chord"

A circle's maximum chord is its diameter, not its radius.



It goes without saying that the radius is not equal to the diameter, since its off by a factor of 2x.

This is important geomtry 101, its not being pedantic.

I don't read that as pedantic. I have no beef whatsoever with what you're saying, it's what we all do. We measure Max Spread. It is simple, and valid. Radii on the other hand is used as predictive tool in probability, as in "Where is the next shot going to land?" That is is the way I understand it. An example:

I shoot a 6 shot group. The order in which they hit the paper is random. In that group there are 4 shots clustered and touching each other very close to the POA. 2 shots each land .6 inches away (radii) from the center of the other 4 shots, and generally in opposite directions. Where will the 7th shot land? Is it more likely or not that it will be closer to the cluster of 4 shots?

Where it lands regardless, it will be measured as a radii from the POA or center of group, expressed in X,Y coordinates.

And our military tests all their weapons using mean radius...........

“The US Military uses the measurement of the mean radius of all the shots in each group as the definition of the angular accuracy of a given firearm, not this ‘extreme shot distance’ measurement,” explains Chris Long, a respected researcher, and writer on precision rifles. Ballistipedia corroborates that saying, “Armed forces also often explicitly uses the more statistically powerful Mean Radius and Circular Error Probable measures.”

There are two very important words in that quote: Angular accuracy. Say I have a very precise rifle, shoots bugholes all day. I proudly display those on the net. But the bugholes are 1 MOA away from the POA. Hmmm.
 
  • Like
Reactions: Marbles
I've seen lots of thought being expressed on this forum about accuracy, how to measure it, and how to rate its importance. There are some competing thoughts. Most of these thoughts deal in numerics. While I like to see a series of nice 1MOA or smaller groups, its not the numbers which are key to me.

I think in terms of the effectiveness of the shot, what it's intended to achieve, and how well it does it. I use the term "defeating" the target. IMHO there's a big difference between 1moa mean radius when you're shooting a mouse, and when you're shooting a moose. One of them is highly likely to "defeat" the target, and the other, maybe less so, maybe a lot less so.

It's an important difference from shooting groups and measuring them in numerics, and I'd like to see how folks here think about that difference. Ii this departure from conventional thinking important or not, and if so, how so?

Greg
 
  • Like
Reactions: JAS-SH
The measured precision and accuracy of the system is used to tell you the odds of defeating any particular target. WEZ analysis tells you what factors are most important in target engagements. On a moose, the target area may be 2 MOA. Is an increase in precision from 1 MOA to .5 a factor if we read wind +/- 5 mph ? How about if the target area is 1 MOA? Do enough of that analysis and you find that in most real world situations precision is rarely the limiting factor if it groups at all. Most factory rifles probably shoot better than the operator, it's odd to find something that can't hold a couple minutes out of the box with whatever is on the ammo shelf.
 
  • Like
Reactions: JAS-SH
Here's another example that relates to my last two posts. This is a different target - 2 MOA with a 1/2 MOA center. Image below is what you see in the software:


i-4w57zMX-L.jpg


And here is the Virtual on it with all the data:
i-3q23VGg-L.jpg


We are all out at the range when I shoot this group. and have all the info above. Where would you place your bet on where the next shot will land?
 
We measure Max Spread. It is simple, and valid. Radii on the other hand is used as predictive tool in probability...

We use a circle to define group size, not a "radius" or a "diameter". The radius/diameter relate to the circle. We use a circle because a the uncertainty of a ballistic trajectory is best thought of as CONE around the flight path, and lo and behold when you projects a cone on a 2D surface you get a circle. :cool:

Once you undestand that, you will get a lot less confused.
 
[cont]

A "radius" is important in the sense that a circle has a radius, so having a radius can allow you to approximate a circle. But having a diameter can allow you to define a radius, since 1/2 a maximal chord is a radius by defintion.

The reason we use the circle in its complete shape to measure accuracy is worth considering: If I am aiming for a target of 1 MOA DIAMETER with binary [0,1] outcome, I don't care if my radius is x or y specific value.

IE, if I want to hit a 1MOA diameter target, I want a measuer of dispersion that indicates a less than 1MOA diameter. In order to do this, you need a dispersion radius that yields a sub-moa dispersion diameter. IE, th is only possible with Dispersion Radius < 0.5MOA
 
Last edited:
Actually, you also require zero aiming error, and a radius/diameter that has the target center and POA center at the same point. It matters not at all if you specify radius or diameter, so long as you also state which you are using. While we need to know the dispersion, the offset from the POA is also important. If you shoot .1 groups that land randomly centered 1 MOA from the POA, that's going to be an issue.
 
I'm not only a photographer, I'm an analyst by profession. I gravitate to hard core analysis when I shoot, even keeping a log of every shot and where it went. Anal is the prefix to analyst so sue me! After reading a a very recent and pretty comprehensive article over at PRB on statistical analysis as it concerns rifle shooting I downloaded the free trial for the OnTarget TDS software today. I really recommend this PRB 3 part series. It is chock full of good information. I also recommend that if you're going to read more than one of the three parts nonstop to get a pot of coffee going! There are a lot of quotes included in there from some of the foremost experts on this subject, including Brian Litz, and so I highly recommend the series.

Putting the PRB three part articles and the OnTarget software together was an eye opener. Luckily, I had saved some of my targets from previous shoots so that I was able to use them for testing and analysis. Having saved some of the targets I've shot with my .22 Magnum trainer I was able to put together a picture that matches the articles to a tee. Makes me a believer, at least until I get additional data but that's not likely to contradict the results.

Here are the link to the first PRB article post. You can find the rest on the follow up links:

https://precisionrifleblog.com/2020/11/29/statistics-for-shooters/

Downloaded the software and got to work. I did not cherry pick my groups. Actually, the opposite is true. I can show you some .4 MOA 5 shot groups at 100 yards but I purposely picked the average/worse groups, so I could see what patterns emerged. The software allows you to superimpose multiple groups which is superb. The end result is that you can shoot different groups any way you want, in my case, 2, 3 shot groups per target and aggregate them over a number of targets into one virtual target. Just happens that this is exactly what I had done. I shot 3 shot groups, reloading in between, 2 groups per target. Most of the time......

All shots were made at 100 yards and aimed at 1.5 MOA targets, with a 1/2 MOA center circle and a 1/4 MOA diamond inside of that. Here is what the target looks like (I print my own):

i-2pMfQhh-X2.jpg


OnTarget has a "virtual" target display that does not include the actual target image. Once the software figures out (with your input) where everything is it generates a virtual target/impacts along with a TON of very meaningful data, including mean radius and CEP (Circular Error Probable) numbers. The US Army uses mean radius exclusively in their evaluation of all weapons. We all use extreme spread, which is good, but not really so much for statistical probability of where your next shot will land. This is because extreme spread weighs outliers in the groups much heavier than the rest of the shots. The original is always available if you need to look at it, but for analysis it becomes superfluous This virtual mode is the mode I used for analysis.

So on to the targets. The targets were made shooting 2 - 3-shot groups into 5 separate targets. The number of shots varied and I actually lost a shot somewhere because the numbers of shots do not match multiples of three by two shots.

This is the first virtual target - 2, 3 shot groups with a LOT of important numbers, including mean radius, SD, and CEP. I wont get into that, you'll have to read the articles :)!

i-7VzxTGn-XL.jpg


But look at what happens when I added 4 more 3 shot groups (losing one shot somewhere) for 17 shots:

i-dRt8nMS-XL.jpg


At this point it looked like I had some fliers. Fliers are what most people don't count. In this image it looks like there at 2 fliers (bottom center left). So do I take them out? Look at what happens when I overlay more groups:

i-sF8zRQx-XL.jpg



Now I'm up to 11-3 shot groups. Lost another shot somewhere (again) so it becomes a 31 shot group. The important point here is that the shots that looked like fliers are not anymore. They are grouped so -- no flyers. By the way the light purple circle is not an MOA or inch measurement, it is SD and CEP related. CEP (Circular Error Probable) is an excellent measure of accuracy and precision. I want more!

All the best,

JAS

So I agree with the statistical algorithms, but hate they can not or will not compile the app for other OS types. I use R to analyze shotgun and rifle patterns most of the time, but I applaud the simple interface of the app. Just what most folks need for analyzing their ammo/rifle combinations. Thanks for posting and Merry Christmas.
 
We use a circle to define group size, not a "radius" or a "diameter". The radius/diameter relate to the circle. We use a circle because a the uncertainty of a ballistic trajectory is best thought of as CONE around the flight path, and lo and behold when you projects a cone on a 2D surface you get a circle. :cool:

Once you undestand that, you will get a lot less confused.

I'm not "confused". Maybe I'm not explaining myself correctly. It happens. See if this matches in any way what you're trying to say:

http://ballistipedia.com/index.php?title=Circular_Error_Probable

"
Circular Error Probable (CEP)

The Circular Error Probable CEP(p) for p∈[0,1) is the estimated radius of the smallest circle that is expected to cover proportion p of the shot group. Some authors restrict the name "CEP" to the case of p=0.5, and refer to, e.g., R95 for p=0.95. We make no such distinction here.

How CEP(p) should be estimated depends on what assumptions are made regarding the distribution of radial errors, i.e., the distribution of miss distances of shots to the point of aim (POA). In turn, the distribution of radial error depends on the bivariate distribution of x- and y-coordinates of the shots. If the x- and y-coordinates of the shots follow a bivariate normal distribution, the radial error around the POA can follow one of several distributions, depending on the cirumstances (Beckmann 1962; 1964): ...."
 
Go to this thread, we are doing a more practical application of this,

 
Go to this thread, we are doing a more practical application of this,
ip

Looked at the thread. I'll play as soon as I can get to the range. I'll chip in on most anything that can help us to become better shooters.

Thanks!