Truth or myth…? I’ve heard the following advice when testing pellets for air rifle.
“For each new batch you test, you need to ’break in’ the new batch for your barrel. So the first group of 10 shots is just for that. The second group of 10 shots will typically be smaller. Use that one in your decision”.
Anybody out there who can confirm this? Or wants to challenge this?
Pellet testing, first and second group: truth or myth?
Moderators: pilkguns, Marcus, m1963, David Levene, Spencer
If you are testing pellets or switching lots and you clean the barrel you will want to shoot a dozen or so shots to season the barrel back with lead. Other than that I can't see a whole lot of reason to do as you mention. There is no such thing as breaking a barrel into a pellet. If the lube is drastically different you might want to clean or shoot a few through it before testing just to get that type of lube smeared in the tube. Maybe that is what you are mentioning. Lot to lot from the same vendor with the same lube, waste of time.
Bo
Bo
*** Warning: statistical content ahead ***
In my latest Quest for the Perfect Pellet, I tested 19 different batches from 1 vendor. For each batch, I did 2 groups of 10 shots.
One rifle (Haemmerli AR50), and simultaneously measured V0 (which was pretty constant, typically the fastest and slowest of 10 shots was within 1 or 2 m/s).
The average shot group diameter was 6.01 mm for the first and 5.85 mm for the second group. Standard deviations were 0.75 and 0.72 mm. So just looking at the averages, the second group was a tad smaller than the first group. But the difference is small, and not statistically significant [t-test: p=0.41].
What surprised me most was the poor correlation between the diameters of the first and second group. The correlation coefficient was merely 0.39, i.e. an r-squared of merely 0.16, and only JUST on the level of marginally significance [p=0.09].
But then again, it has been stated before (oh, wait.. I wrote that myself ) that many, many shots are needed if you want to distinguish reliably between 2 batches.
In my latest Quest for the Perfect Pellet, I tested 19 different batches from 1 vendor. For each batch, I did 2 groups of 10 shots.
One rifle (Haemmerli AR50), and simultaneously measured V0 (which was pretty constant, typically the fastest and slowest of 10 shots was within 1 or 2 m/s).
The average shot group diameter was 6.01 mm for the first and 5.85 mm for the second group. Standard deviations were 0.75 and 0.72 mm. So just looking at the averages, the second group was a tad smaller than the first group. But the difference is small, and not statistically significant [t-test: p=0.41].
What surprised me most was the poor correlation between the diameters of the first and second group. The correlation coefficient was merely 0.39, i.e. an r-squared of merely 0.16, and only JUST on the level of marginally significance [p=0.09].
But then again, it has been stated before (oh, wait.. I wrote that myself ) that many, many shots are needed if you want to distinguish reliably between 2 batches.
I LOVE statistics. Why, I even read a book entitled, "How to Lie With Statistics".
It didn't help much with pellet testing, but was certainly useful in figuring out the political commentaries and when I did some government work as a writer.
Fortunately, I'm an air pistol shooter so I don't get quite so anal about test groups, but when I test, I shoot 25 or 30 shot groups at one time. That certainly separates out MY choices.
It didn't help much with pellet testing, but was certainly useful in figuring out the political commentaries and when I did some government work as a writer.
Fortunately, I'm an air pistol shooter so I don't get quite so anal about test groups, but when I test, I shoot 25 or 30 shot groups at one time. That certainly separates out MY choices.
Grzegorz
You missed the interesting bit. Jeroen sampled 19 different pellet batches. The two averages are the average of the first 10 pellets of the 19 groups, and the second is the average of the second 10 pellets of the 19 groups.
A tad more than "2 lots of 10 shots". So, yes, I think Jereon is saying there isn't much between them. I wouldn't bet my house that the second 10 shots would be better than the first.
Jeroen: As a matter of interest, of the two 10 shot groups, how many times was the first group bigger than the second??? I just bet my house I think......
You missed the interesting bit. Jeroen sampled 19 different pellet batches. The two averages are the average of the first 10 pellets of the 19 groups, and the second is the average of the second 10 pellets of the 19 groups.
A tad more than "2 lots of 10 shots". So, yes, I think Jereon is saying there isn't much between them. I wouldn't bet my house that the second 10 shots would be better than the first.
Jeroen: As a matter of interest, of the two 10 shot groups, how many times was the first group bigger than the second??? I just bet my house I think......
I have not missed that. I am just not so enthusiastic about this averaging over 19 DIFFERENT batches. That's all. Anyway, I like such tests very much, but very often there is a lack of statistically sufficient material behind them and that is a problem (exemplary, the common use of "a diameter" that in fact reflects a distance of only two the most distant shots in a group, etc). Just my 2 cents in this soft subject :-)RossM wrote:Grzegorz
You missed the interesting bit. Jeroen sampled 19 different pellet batches. The two averages are the average of the first 10 pellets of the 19 groups, and the second is the average of the second 10 pellets of the 19 groups.
...
Been a while since I studied statistics, (about 3 decades) but it has been my experience that group shape is much more important than group size. Now we all know that size DOES matter in all things but I would much rather have a round group with consistently distributed shots than one with a smaller size and irregular shape.
Also maybe spending time practicing instead of testing pellets would be more beneficial.
Also maybe spending time practicing instead of testing pellets would be more beneficial.
@Grzegorz: in the statistical test, I am not averaging over 19 different batches. For Batch 1 to Batch 19, I compared a first with a second 10-shot group.
I am well aware that "group diameter" is not the best measure. But what can you do if you've only got a clamp and a conventional paper target @ 10m (and no electronic target that gives the locations of each individual shot?). Group shape as gwsb suggested, yes, but how can we quantify that?
If you think 19 batches are not enough: how many would be sufficient (and why?)
Of course, spending time at training is preferable over spending time at testing pellets. I just want to make sure that _I_ am the limiting factor, and not my pellets.
I am well aware that "group diameter" is not the best measure. But what can you do if you've only got a clamp and a conventional paper target @ 10m (and no electronic target that gives the locations of each individual shot?). Group shape as gwsb suggested, yes, but how can we quantify that?
If you think 19 batches are not enough: how many would be sufficient (and why?)
Of course, spending time at training is preferable over spending time at testing pellets. I just want to make sure that _I_ am the limiting factor, and not my pellets.
@JeroenH Yeah, that is what I have tried to express - we have usually a problem with a sufficient statistical probe. Sorry, from your post I understood you averaged over 19 batches. Anyway the problem is still present... What to do? I test ammo in different ways still looking for the most efficient procedure (efficient - read: lowest cost possible, maximized result, acceptable statistically probes). One of the methods I have found as working quite well is based on the surface measurements expressed in pixels. So, simply I shot 10 shot series (four for each ammo tested) and I collect 4 groups plus (very important in my opinion) one 40 shots group. Then I scan these groups and I use PSE11 soft to count pixels for each group image. I count internal surface and external surface as well (so somehow, a shape is taken into account this way). Then I compare results. See below how these two (internal and external) surfaces are selected in PSE. Maybe this works for you, too?
The report from one of such tests is given here in .... Polish :-) but take it easy, I have added comments in frames in English :-)
http://www.lfits.pl/Ammotesting.pdf
PS. This example concerns .22'', but...
The report from one of such tests is given here in .... Polish :-) but take it easy, I have added comments in frames in English :-)
http://www.lfits.pl/Ammotesting.pdf
PS. This example concerns .22'', but...
Last edited by Grzegorz on Wed May 22, 2013 12:59 am, edited 1 time in total.
-
Metookevin
- Posts: 27
- Joined: Wed Apr 03, 2013 7:27 pm
- Location: Sydney Australia
What about velocity? You need to consider velocity variable on pellets. In getting my air rifle "calibrated" for the current stock of pellets I set up in machine vise minus stock which gives easy access to velocity adjustment screw on fwb P70 or P700. Then x- check velocity adjustment with group size with mount chronograph. With electronic targets I will fire 60 shots which gives good indication once I have done preliminary 20 shot groups to narrow down a desirable velocity.
The machine vise rests on an adjustable angle plate with improvised jack screw for vertical aim. Clamps are used on existing bench top to aim horizontally. With this rig I can aim into centre of electronic target to get meaningful results.
On a good day I will shoot 584 in training with tin rws r10 (rws yellow tin does not shoot as well in my group tests ) on my feet. I suspect that lower velocity may be easier to control off the shoulder even though group would be larger but that is only a thought and have not tested this idea.
I have also noted that all things being equal that group size is affected by dirty barrel when doing these group tests.
The machine vise rests on an adjustable angle plate with improvised jack screw for vertical aim. Clamps are used on existing bench top to aim horizontally. With this rig I can aim into centre of electronic target to get meaningful results.
On a good day I will shoot 584 in training with tin rws r10 (rws yellow tin does not shoot as well in my group tests ) on my feet. I suspect that lower velocity may be easier to control off the shoulder even though group would be larger but that is only a thought and have not tested this idea.
I have also noted that all things being equal that group size is affected by dirty barrel when doing these group tests.
Grzegorz, thanks for sharing your approach, and the scan.
Actually, I made such scans of my air rifle targets myself as well recently. Only, instead of PSE, I used a Matlab programme to measure the 'hole' (internal surface only, in your terminology).
See an example below: left = original scan; middle = scan with identified hole filled; right is extracted hole with largest distance.
In this way, I can easily quantify the hole diameter, but also the hole area.
Or other measures that can be derieved automatically (and perhaps extress the shape of the overall hole?)
Why? Well, for example to look at effect of pellet caliber on the group size. I had 48 batches of RWS R10. Results: see the figure below. And yes, this effect was statistically significant. Telling me for the next time that I'd better start my search among the 4.48 pellets than among the 4.50 ones. So hopefully making my future testing more efficient.
Actually, I made such scans of my air rifle targets myself as well recently. Only, instead of PSE, I used a Matlab programme to measure the 'hole' (internal surface only, in your terminology).
See an example below: left = original scan; middle = scan with identified hole filled; right is extracted hole with largest distance.
In this way, I can easily quantify the hole diameter, but also the hole area.
Or other measures that can be derieved automatically (and perhaps extress the shape of the overall hole?)
Why? Well, for example to look at effect of pellet caliber on the group size. I had 48 batches of RWS R10. Results: see the figure below. And yes, this effect was statistically significant. Telling me for the next time that I'd better start my search among the 4.48 pellets than among the 4.50 ones. So hopefully making my future testing more efficient.
- Attachments
-
- TargetProcessing.JPG (18.44 KiB) Viewed 3241 times
-
- Group diameter as a function of caliber (Mean and 95% Conficence Interval)
Nice to see the seriously treated experiment. Like it very much. Surely Matlab offers more possibilities than PSE :-) I am only affried that admin will kick off us from this subject soon, maybe it would be fine to renew any older one devoted to ammo testing? But that is not me, I have to switch off my PC and start to manage my luggage as tomorrow I fly to Munich and will be busy with supervising scoring procedures rather than ammo tests :-) However, if I convince any shooter to share the ammo test procedure with me - gonna post it after (if permitted, of course!)
Grzegorz, have a good trip and of course I'd be interested to hear about procedures from others.
Rover, group sizes incerase with pellet size for my rifle. For other rifles it may be different.
Rover, group sizes incerase with pellet size for my rifle. For other rifles it may be different.
Last edited by JeroenH on Thu May 23, 2013 6:42 am, edited 1 time in total.
(Sorry I missed your question earlier, Ross).RossM wrote: Jeroen: As a matter of interest, of the two 10 shot groups, how many times was the first group bigger than the second??? I just bet my house I think......
The second 10-shot group was smaller than the first 10-shot group in 12 out of 19 batches (63%). Which seems to suggest that the second groups is indeed smaller than the first. But a sign test shows that this is not a significant effect [p=0.35]. So, you get to keep your house.