Testing Ammo

[JerrySharrett]

Landy, thanks for the effort and may your ears (corn) dry without mold and may your tooth dentist be in his best scrubs.

.

 

[HuskerP7M8]

Do you know the worst things about your posts? There are simply not enough of them. And, they get lost in the masses of other posts.
I'd suggest to the forum administrators that all of your posts be made into "stickies" so we wouldn't have to hunt them down.

LOL....Ron, I'm going to pray this thread is so confusing that Wilbur deletes it! But I do really appreciate what you said.

Landy,
I'm not even certain how to ask this question. And I suspect that the answer would be merely a judgement call rather than a statistically valid one. But, let's try.
Let's say you had equipment that you felt would be competitive at the national level (would give you a reasonable chance of being in the top 10 at the ARA Nationals).
How many 5-shot groups (measured using ATC) would you need to shoot and what would the ATC need to average to make you comfortable enough to buy a few cases and shoot it at national competitions?
How's that for a totally unscientific question????????? Does the question even make sense?

Hawkeye Wizard

The question makes perfect sense, but answering it with a high level of confidence is hindered by a few problems.

You'd need enough data on the rifle/ammo combination to reach a confidence level that's acceptable to quantify the precision of your rifle. Essentially that wouldn't be much different than the procedures used to determine acceptable confidence levels for selecting ammo based on your lot testing.

I'm going to cheat in order to save my time by linking to an excellent article written by Dr. Geoffrey Kolbe. I would suggest you completely ignore the equations and charts presented in the article and focus on his general comments. I'm hoping this will give everyone a general idea of what statistical confidence levels are and the size of the data set to achieve them.
I have to admit I was more than a little depressed many years ago when I first realized the sheer magnitude of data required for high levels of confidence.

http://www.geoffrey-kolbe.com/articles/rimfire_accuracy/group_statistics.htm

Assuming you've been successful in gathering the data to evaluate your rifle's performance, how do you compare it to the top ten finishers at the National Matches?
When I've done this in the past I used the scores from the Indoor Nationals, the large Indoor Tournaments, and the ARA aggs for the Indoor season.
Maybe there's a way to use the outdoor results to make comparisons, but I felt the results of such an analysis would skew the results away from equipment/ammo and too much toward a competitor's skill level doping the wind.

Because I'm a data junkie I know within a fairly narrow range the performance potential of my equipment and I've been very successful in predicting where I'll finish in the large indoor matches.
I've also been somewhat successful at predicting where I'll finish in large outdoor matches, but only when the conditions are such that I'm able to narrow the gap between my skills in reading conditions and say maybe the top twenty or thirty shooters who are nearly always hanging at the top regardless of conditions.

As regards any hard numbers for avg extreme spreads for 5-shot groups and/or the Mean Radius for large shot distributions, I've set some personal benchmarks that seem to be fairly accurate.
By my best estimates, ATC or MR needs to be at least 0.100" and 5-shot grp avg's need to be at least 0.25". This should put you somewhere around the 60 or 70 percentile range (a finish of about 30th to 40th in a 100 competitor match).
To reach the 90th percentile, I estimate somewhere in the low to mid 0.8's for ATC/MR and low 0.2's for 5-shot avg's.

I would caution everyone that my estimates for 5-shot grp avg's are based on "CORRECT" measurements, and not those you'll commonly see reported on internet forums.

Landy

 

[HuskerP7M8]

Landy, thanks for the effort and may your ears (corn) dry without mold and may your tooth dentist be in his best scrubs.

.

Thanks Jerry,

Corn harvest was finished last night and I'd be a happy camper, but the Dentist lied to me about 6 to 8 hours of discomfort today.
I may shoot myself after I'm thru posting to end the misery!

Landy

 

[HuskerP7M8]

Hawkeye Wizard
Thanks. I overlooked Landy's explanation that ATC = Mean Radius. But that brings back up the question I thought I had answered in my incorrect "explanation" of why ATC is better than group size.

Landy
Why is ATC better than the Group Offset measurement in TDS? If I understand Group Offset, which I may not, it is measuring the distance of the group from center which is the desired point of impact. With ATC being the mean radius *of the group* then your group conceivably could be very small but significantly away from the center of the target and thus score poorly.

Well, to answer my own question, group offset could have a similar flaw. The group could be very well centered on the target but be a very large group and not score well.

It seems to me a combination of ATC and group offset would be most valid. Have you looked at anything like that?

Tom

Tom,

ATC/MR and Group Offset are both useful tools, but the ATC/MR is used to determine precision and the Group Offset is primarily used to determine accuracy. It's important to remember there's a distinct difference between the two even though they're often used interchangeably in informal discussions.
Rather than taking the time to explain it, might I suggest you revisit the thread I linked to in post #16. There's some discussion within it that may be helpful concerning "Accuracy" and "Precision" and it again saves me some time.

Landy

 

[HuskerP7M8]

Is the third regression (average distance to center vs score) measuring mean POI to center of bull for 25 bulls, then regressing this against that target's score? If so the high correlation makes perfect sense: The smaller the average distance from POI to bull center, the higher the score (score is inversely correlated with mean POI-center distance).

Hi Kim,

It's obvious you're familiar with some of the terms I've used and for all I know may be more skilled in using them than I am. All of you need to know/remember I'm a self-taught "expurt" in these matters.
Anyway, if I'm interpreting correctly what you said....yes, you're correct. And, if I'm correctly interpreting your comment on "Inverse Correlations", that's only because of the way I created the charts so they show the same statistical "slope". I thought it might be easier to compare the charts with a "slope" that's uniform from chart to chart.
Let me know if I'm correct or need a Rosetta Stone. LOL

The other two regressions say 2 or 5 shot groups. Are these groups "virtual"? ie sampled from scans of scored targets? If so how are the groups sampled from the targets? Consecutive shots?

Thanks

Kim

Yes, they're consecutive shots and not a sampling or an Excel Monte Carlo simulation. That being said, there are some advantages to using random samplings but I think this thread is confusing enough as is.

Landy

 

[KimZ]

Stats

Thanks Landy. I studied statistics and agree measurements and stats are useful tools to assess accuracy.

You know this, but for others interested the sign of regression slopes (correlation) come from the relationship between measured variables. Positive correlation is when two variables scale together - and the slope of the regression line is positive. For example, weight and height in humans. Negative correlation is when they scale inversely: the smaller the miss the higher the card score.

The correlation coefficient (R) is a measure of how tightly the data points cluster around the regression line. (Correlation coefficient is not the same as regression slope, but they are the same sign) Perfect correlation is 1.0 (all points exactly on the regression line). Low or un-correlation is closer to zero. For most real-life data, the cloud of points can look quite random but least-squares regressions can still show significant correlation (statistically significant means the observed correlation has a low probability of occurring just by chance).

The reason your R values are high is that you are correlating two related (non-independent) measurements: card score vs. group size (within the same cards). But this is useful since you were trying to demonstrate that scores are better for ammo that groups small. The caveat would be that 2 or 5 shot groups could be tight - but it is possible that tight groups at the beginning and end of the card could have a shifted average POI. So the most useful groups would be at least as large as the number of bulls shot. This could be assessed by looking at miss measurements (POI to center) from first to last bull shot (assuming you shoot in the same order every time), over many cards.

There may be other interesting information in the scanned targets. Have you looked at the direction of misses (POI to center)? ie, high, low, right, left. There could be patterns due to ammo factors, barrel temperature, change in miss direction over time, etc. Could also be interesting to check vertical miss vs chrony FPS, etc.

Data can also show psychological effects. I tabulated a number of cards from my air rifle BR matches. I always shoot the same row-order - and found most misses occurred in the last row shot. This confirmed that I tend to rush the last row to finish the card....and finish the tension. Now my goal is to chill on the last row )chill(

Kim

 

[HuskerP7M8]

There may be other interesting information in the scanned targets. Have you looked at the direction of misses (POI to center)? ie, high, low, right, left. There could be patterns due to ammo factors, barrel temperature, change in miss direction over time, etc. Could also be interesting to check vertical miss vs chrony FPS, etc.


Kim

Kim,

There's not much I don't track and document when testing, including everything you mentioned above.
Not all of it is useful, but it's always interesting. I'll show a few screenshots from my spreadsheets, but it might take me a couple of days to get to it.

Landy

 

[TangoTwo]

10-26-2016, 10:10 PM #42




HuskerP7M8



Join Date:Jul 2007Location:NebraskaPosts:344



Assuming you've been successful in gathering the data to evaluate your rifle's performance, how do you compare it to the top ten finishers at the National Matches?
When I've done this in the past I used the scores from the Indoor Nationals, the large Indoor Tournaments, and the ARA aggs for the Indoor season.
Maybe there's a way to use the outdoor results to make comparisons, but I felt the results of such an analysis would skew the results away from equipment/ammo and too much toward a competitor's skill level doping the wind.

Because I'm a data junkie I know within a fairly narrow range the performance potential of my equipment and I've been very successful in predicting where I'll finish in the large indoor matches.
I've also been somewhat successful at predicting where I'll finish in large outdoor matches, but only when the conditions are such that I'm able to narrow the gap between my skills in reading conditions and say maybe the top twenty or thirty shooters who are nearly always hanging at the top regardless of conditions.

.

Landy

[END QUOTE]


Very astute observation concerning shooter skill levels. Back when Bob Holbruner and Harry Deneen were shooting they called calm days trigger pulling contests. When the wind blew is when their talent really shined. To base results on their talent versus most shooters, would truly skew the results.

Ken Henderson

 

[tim]

[END QUOTE]


Very astute observation concerning shooter skill levels. Back when Bob Holbruner and Harry Deneen were shooting they called calm days trigger pulling contests. When the wind blew is when their talent really shined. To base results on their talent versus most shooters, would truly skew the results.

Ken Henderson

Always kind of nice to see "shooter skill levels" comes in handy now and then. So those famous WLM rifles don't actually shoot themselves after all? Very refreshing.

 

[ahighe]

Methodology?

Real world target scores? That reminds me of some work I've done in the last month or two and I think you and others may find it interesting. It should also assist everyone in setting benchmarks based on groups shot during your testing if your preference is to continue shooting groups rather than buying software and working with statistics you're not familiar with.


All 3 charts use the same data from a single barrel utilizing the same 3,375 shots. I didn't look, but there's probably some 30 to 40 different lots of ammo in the data.
The reason I mention the number of lots is because I believe that may be the best way to make a determination of whether or not you should give up on a barrel and/or build and start anew. If after I've tested some 20+ lots and the charts I'm presenting suggest none of the results are competitive, I'll rebarrel or start a new build. If I sell the rifle I've given up on, I'll either rebarrel it before I sell it or tell the new owner they might be better off rebarreling it because the odds are they won't be happy with it. I don't want to lay awake at night thinking I'm guilty of selling junk even if I take a hit financially.

All 3 charts show the relationship or statistical correlation to real world scores (ARA Score/C'd) for testing results using 2-shot groups, 5-shot groups, and ATC (Average To Center) also known as MR (Mean Radius).

The red line thru the data points is a trend line and in this case it's a linear trend line. Excel calculates the best fit thru the data points and creates the line.

R2 is a calculation Excel uses to tell us how well the data fits the trend line. Higher numbers are better.

Correl is the term used in the Excel formula bar and is short for correlation. Whether minus or positive, the higher the number the closer the relationship or the higher the correlation is to predicted scores.
You should notice there is some scatter that prevents you from drawing an intersecting line from the group size average in the horizontal axis of the charts and predict an exact score. This scatter is perfectly normal and can only be lessened with the addition of more data. Considering this data is composed of 135 separate tests and 3,375 total rds, it should be obvious why scores vary so much from target to target even under benign conditions with the same rifle and ammo.

The other important result to notice is how the prediction for real world scores suffers if for some reason you were to decide to use 2-shot groups in your testing. The same happens if you use anything other than groups with 5 to 6 shots in them. Testing using 2-shot, 3-shot, 10-shot groups, etc don't work well when testing ammo or evaluating rifle performance. Again, this is well known but probably unfamiliar to most of you.

There's a lot to digest here and I'm certain there will be disagreement. All I can do is present the data and let everyone draw their own conclusions.

I'd also like some input as to how understandable this is. I don't care if anyone disagrees with what I've said....just want to know how many of you can make sense of it. I'd really appreciate the input.

Thanks,
Landy

[/IMG]

Hi Landy,
It's not clear to me how the values for each point are calculated. For example, is the ARA score and ATC for 25, 50 points and use the same data or something else? Are the values for 2, 3, and 10 shot groups averaged subsets of the same data?
Thanks.
Albert

 

[HuskerP7M8]

I haven't forgot about this thread, but had a little relapse in my dental condition and developed an infection. Don't feel too well and on a heavy dose of antibiotics. Plus I have visitors scheduled to lot test ammo all day.
Will make a few more comments but don't know when.

Landy

 

[HuskerP7M8]

I think this thread has run its course, but I told Kim I'd post a couple of screenshots and I wanted to complete that before I bow out.

There may be other interesting information in the scanned targets. Have you looked at the direction of misses (POI to center)? ie, high, low, right, left. There could be patterns due to ammo factors, barrel temperature, change in miss direction over time, etc. Could also be interesting to check vertical miss vs chrony FPS, etc.


Kim

Kim,

These are just some visuals that my spreadsheets create automatically after the point data from TDS and the velocity string from the chronograph are input. I like to be able to quickly look at charts and get a general impression of how the testing went and possibly see if there's any potential problems.
Absent the charts, there are many and some better ways to work with only the numbers that include everything you mentioned.

Landy

 

[wicklowman]

Here's how I look at it. Lots of ammo will shoot several groups well. By shooting a whole card for score you're shooting a 25 shot group. That's why I'll shoot the whole card. That's when ugliness rears it head.
Early on I would buy ammo that grouped well. Bit me in the ass many many times. Oh did it bite me.
That's why Eley must shoot 40 shot groups in they're testing. To me its the same thing. Pretty hard to shoot a 40 shout group without being in a fixture,


Keith

Hi keith,You don't shoot a forty shot group at the Eley test centre here in the UK,you shoot 4 10 shot group which are then imposed one on top of the other by computor,this then gives you the overall group size for forty shots.I have had groups measuring from 17mm to 25 mm.I shoot a few groups of 5 shots just to see what consistency I have in the conditions then just shoot a card,if I am holding the 10 or a smidgen off the I am happy with with my ammo.John F.

 

[linekin]

When Ken's wind flag tips over my shots land outside the scoring rings......I learned that the other day!!!!
Not the best idea I've had testing in those conditions.

Keith