[Dr. Isotope]

I like these tests. They're strong in spirit, but in the end they're essentially modern art: interesting, but generally indicative of nothing. Due to the nature of the paintball as a projectile (an imperfect, flexible, seamed-sphere filled with a viscous fluid) barrel accuracy tests are impossible. The Punkworks fellas are doing paint tests. Change the paint, and every single data point will likely change along with it. I feel bad for the folks that put any stock in (or money spent as a result of) the results, unless they happen to use those particular brands of paint.

[Leftystrikesback]

Leafy, on Mar 22 2009, 09:35 PM, said:

cockerpunk, on Mar 23 2009, 12:28 AM, said:

Leafy, on Mar 23 2009, 12:28 AM, said:

I'm just confused that it looks to me that the 21 inch barrel was the most accurate and that goes against everyother test i've ever seen.

you taken stats yet leafy?

no next year. I'm in the joke of a class called linear algebra right now.

Take that seriously... Linear algebras is one of the most important things to know if you're going to do any kind of data manipulation on a computer. You won't learn this for a while but most math problems can be represented using matrices.

Both the 21 inch 2 piece and th 10 inch 1 piece have the smallest vector, which would imply they are the most accurate, however if you look at the "Calcs" page you can see they have the highest residuals, which is basically an indicator of how random the samples are, and how well the statistics can be matched with reality.

Above the residual values it says "lower is better" which is incorrect, he means "closer to zero is better" because it implies a more random sample of the population.

The things I take from this are the trends:

Large overbore and small underbore are both better than bore matching or slight overbore

Longer barrel seems to show less accuracy. A couple inconsistencies in that trend though (the aforementioned 21" barrel)

Quote

I like these tests. They're strong in spirit, but in the end they're essentially modern art: interesting, but generally indicative of nothing. Due to the nature of the paintball as a projectile (an imperfect, flexible, seamed-sphere filled with a viscous fluid) barrel accuracy tests are impossible. The Punkworks fellas are doing paint tests. Change the paint, and every single data point will likely change along with it. I feel bad for the folks that put any stock in (or money spent as a result of) the results, unless they happen to use those particular brands of paint.

Data points will always change between tests whether the paint is changed or not. Trends are much less likely to.

This post has been edited by Leftystrikesback: 23 March 2009 - 02:23 AM

[Leafy]

I thought it was supposed to say lowest absolute value was better on that. Thanks for clearing that up.

[brycelarson]

Leftystrikesback, on Mar 23 2009, 02:20 AM, said:

Above the residual values it says "lower is better" which is incorrect, he means "closer to zero is better" because it implies a more random sample of the population.

Leafy, on Mar 23 2009, 06:42 AM, said:

I thought it was supposed to say lowest absolute value was better on that. Thanks for clearing that up.

no, we really mean "lower is better" - a negative number shows that the vector for that particular barrel was better (lower) than the mean vector. Now, the columns on the right are the 99 and 90% confidence interval. All of the vector residuals are inside the 99% AND the 90% CI.

This means to me that any attempt to try to say ______ is better than ______ on this test are potentially dangerous. I think that we've shown again that a tube is a tube.

[cockerpunk]

Dr. Isotope, on Mar 23 2009, 02:54 AM, said:

I like these tests. They're strong in spirit, but in the end they're essentially modern art: interesting, but generally indicative of nothing. Due to the nature of the paintball as a projectile (an imperfect, flexible, seamed-sphere filled with a viscous fluid) barrel accuracy tests are impossible. The Punkworks fellas are doing paint tests. Change the paint, and every single data point will likely change along with it. I feel bad for the folks that put any stock in (or money spent as a result of) the results, unless they happen to use those particular brands of paint.

i like where you were going, but then your ended up some place bad.

what does the bolded sentence tell us about what is most important to paintball accuracy ...

[Toriphilewill]

Has PunkWorks thought of expressing these experiments in a formal paper format? That would remove some of the repeat questions this thread already has regarding what the purpose of this experiment was, and the materials used. If you spend a great deal of time making sure the experiment is controlled as best as possible, then just as much time should be spent explaining them. This should not be taken as an insult or criticism, just a suggestion.

I shoot most of my targets at the 125' and beyond range, so this topic is very interesting to me. I've also calculated the SD, mean, and p-values myself, and can agree with your mathematical results. When I get home I will make a scatter plot to help me visualize the data in another way.

[Leftystrikesback]

brycelarson, on Mar 23 2009, 06:48 AM, said:

no, we really mean "lower is better" - a negative number shows that the vector for that particular barrel was better (lower) than the mean vector. Now, the columns on the right are the 99 and 90% confidence interval. All of the vector residuals are inside the 99% AND the 90% CI.

This means to me that any attempt to try to say ______ is better than ______ on this test are potentially dangerous. I think that we've shown again that a tube is a tube.

I see, for some reason I thought you were comparing the individual data points from the barrel to the barrel's own vector (this is what I think of as residuals)... and if they didn't add up to zero that would not be good heh. Should probably stop trying to do math late at night.

I'm going to play devil's advocate and say a spread of 9 inches at 125 feet is significantly better than a spread of 13 inches. The area covered by a circle with diameter of 13 inches is more than twice that of a circle with a diameter of 9 inches. So why this significant difference in some barrels (with the same bore size!)?

This post has been edited by Leftystrikesback: 23 March 2009 - 01:01 PM

[Spitlebug]

Here is what I interpret from the 21" barrel being somewhat of a red herring (I can't watch the video because I am at work).

1.) It has everything to do with the spatial length of the barrel to the target. A 12" barrel will be 9" shorter from the target.

2.) If you look at this in terms of exit trajectory a shorter barrel will have a trajectory deflection farther back than the 21" barrel.

3.) A shorter deflection point translates into greater travel on an angular path than the longer barrel. Looking at pure vectors, it is understandable that a ball travelling at the same vector from a shorter barrel will produce a larger scatter pattern.

This post has been edited by Spitlebug: 23 March 2009 - 02:59 PM

[cockerpunk]

Leftystrikesback, on Mar 23 2009, 01:00 PM, said:

brycelarson, on Mar 23 2009, 06:48 AM, said:

no, we really mean "lower is better" - a negative number shows that the vector for that particular barrel was better (lower) than the mean vector. Now, the columns on the right are the 99 and 90% confidence interval. All of the vector residuals are inside the 99% AND the 90% CI.

This means to me that any attempt to try to say ______ is better than ______ on this test are potentially dangerous. I think that we've shown again that a tube is a tube.

I see, for some reason I thought you were comparing the individual data points from the barrel to the barrel's own vector (this is what I think of as residuals)... and if they didn't add up to zero that would not be good heh. Should probably stop trying to do math late at night.

I'm going to play devil's advocate and say a spread of 9 inches at 125 feet is significantly better than a spread of 13 inches. The area covered by a circle with diameter of 13 inches is more than twice that of a circle with a diameter of 9 inches. So why this significant difference in some barrels (with the same bore size!)?

if that is true lefty we should be able to find a trend. is there a pure length trend? a bore size trend? a one piece or two piece trend? a chrono trend?

have you guys by any chance flipped the page and looked at the one pieces?

This post has been edited by cockerpunk: 23 March 2009 - 03:03 PM

[Leftystrikesback]

I already posted the trends that I saw.
The 21" was a notable exception to the trend, which is interesting because then it's accuracy can't be explained by the trends, hence my previous post.

You can't compare the one pieces to the two pieces because different bores and lengths were used. You can't know if trends due to those are skewing your comparison.

[cockerpunk]

Leftystrikesback, on Mar 23 2009, 03:23 PM, said:

I already posted the trends that I saw.
The 21" was a notable exception to the trend, which is interesting because then it's accuracy can't be explained by the trends, hence my previous post.

You can't compare the one pieces to the two pieces because different bores and lengths were used. You can't know if trends due to those are skewing your comparison.

right, what are those trends? thats the question im asking.

if length is a trend then we should be able to say that all the long barrels shot better, or that the longer the barrel, the better the vector. same with everything else. do you notice any of those?

This post has been edited by cockerpunk: 23 March 2009 - 03:29 PM

[Leftystrikesback]

Every barrel shorter than 12 inches shot better than the corresponding barrel longer than 12 inches, in both 1 and 2 pieces. I'm not counting the tribal/ stock cocker barrels since they are different brands. The notable exception is the 21 inch barrel.

Spittlebug is attempting to explain that by saying that the ball exits the barrel closer to the target in the 21 inch barrel, and this means shorter distance and flatter travel overall. However I have a hard time believing that over 125 feet, a 6 inch difference in distance is going to make a noticeable difference in spread. Maybe I'll look at the math more later.

[brycelarson]

we have had a conversation about laying out a standardized report form - it's in the works, but doesn't exist yet.

[Lord Odin]

Spitlebug, on Mar 23 2009, 02:59 PM, said:

Here is what I interpret from the 21" barrel being somewhat of a red herring (I can't watch the video because I am at work).

1.) It has everything to do with the spatial length of the barrel to the target. A 12" barrel will be 9" shorter from the target.

2.) If you look at this in terms of exit trajectory a shorter barrel will have a trajectory deflection farther back than the 21" barrel.

3.) A shorter deflection point translates into greater travel on an angular path than the longer barrel. Looking at pure vectors, it is understandable that a ball travelling at the same vector from a shorter barrel will produce a larger scatter pattern.

You are correct in that the ball's exiting the barrel closer to the target (assuming the gun position remains constant) and it should equate to better accuracy. However, according to this data, that's not the case. If it were, then there should have been a trend of longer barrels getting more accurate.

[Timmy]

i've been sitting here, trying to do the math. my numbers just aren't working out. Maybe it my level of eductaion(grade10) maybe my fatigue. But i just can't get these numbers sraight....

[cockerpunk]

Lord Odin, on Mar 23 2009, 04:28 PM, said:

Spitlebug, on Mar 23 2009, 02:59 PM, said:

Here is what I interpret from the 21" barrel being somewhat of a red herring (I can't watch the video because I am at work).

1.) It has everything to do with the spatial length of the barrel to the target. A 12" barrel will be 9" shorter from the target.

2.) If you look at this in terms of exit trajectory a shorter barrel will have a trajectory deflection farther back than the 21" barrel.

3.) A shorter deflection point translates into greater travel on an angular path than the longer barrel. Looking at pure vectors, it is understandable that a ball travelling at the same vector from a shorter barrel will produce a larger scatter pattern.

You are correct in that the ball's exiting the barrel closer to the target (assuming the gun position remains constant) and it should equate to better accuracy. However, according to this data, that's not the case. If it were, then there should have been a trend of longer barrels getting more accurate.

right, which we dont see.

[CrazyLittle]

cockerpunk, on Mar 23 2009, 01:26 PM, said:

if length is a trend then we should be able to say that all the long barrels shot better, or that the longer the barrel, the better the vector. same with everything else. do you notice any of those?

Well, as length approaches infinity, yes, the longer barrel will be more accurate... Somewhere right around 124'

This post has been edited by CrazyLittle: 23 March 2009 - 05:11 PM

[Leafy]

CrazyLittle, on Mar 23 2009, 06:11 PM, said:

cockerpunk, on Mar 23 2009, 01:26 PM, said:

if length is a trend then we should be able to say that all the long barrels shot better, or that the longer the barrel, the better the vector. same with everything else. do you notice any of those?

Well, as length approaches infinity, yes, the longer barrel will be more accurate... Somewhere right around 124'

but what makes the best effective length on that?

[cockerpunk]

Leafy, on Mar 23 2009, 06:12 PM, said:

CrazyLittle, on Mar 23 2009, 06:11 PM, said:

cockerpunk, on Mar 23 2009, 01:26 PM, said:

if length is a trend then we should be able to say that all the long barrels shot better, or that the longer the barrel, the better the vector. same with everything else. do you notice any of those?

Well, as length approaches infinity, yes, the longer barrel will be more accurate... Somewhere right around 124'

but what makes the best effective length on that?

haha, well for accuracy we should use barrels more then 100 feet long.

ill have to add that to my quote book which already includes "why even have a barrel at all, it just causes barrel breaks!"

[Chris Logan]

Something that cockerpunk has been trying to point out is that a 10" one piece barrel had almost the same vector as the 21". Too little of a difference, in fact, to consider it different at all with the scope of the test.

Cockerpunk: Did you test using DXS frostbite on the one piece and DXS Silver on the two piece? Or is that a typo?

I firmly believe that there needs to be a testing standard, and this isn't it. By using paintballs at all, you are introducing far too many variables into a BARREL test. And then what about a the gun? The gun itself has it's own set of inconsistencies.



If barrel testing was done with a standard tool, at some point you can estimate how inconsistent the tool is. Same with projectiles. Remember the hard plastic target rounds that were available before all manners of foam balls? Those may be a good candidate. Extremely consistent in size. There is a seam. I'm not sure about how close in weight they are to the standard for paintballs.

Punkworks does an amazing job at testing these things. But the tests need to be totally repeatable. So that someone in Canada can set up the same test with a new barrel, and add to a large dataset. So that a new manufacturer can advertise and verify on this scale.

Here was the original question: What makes one paintball gun barrel better than the next?

And we still don't know.