Bought my first chrono this week, M1 Chrony. Went to the range today and tried it out but came away with a question. I have attached a PDF file of the data.

Would the cold barrel overall average fps be a good bit lower than fps after several shots? I ran two strings of 10 through my Vaquero 5.5," one at the start and one later." The second string avg fps was about 40 fps faster than the first with the same load. Is this normal?

Also what high/low spread between 10 shots would one expect to be typical of 45 Colt reloads? As the data sheet indicates the load is VV-N320 7.3 grs with a Hornady 200 gr SWC. The spread made me think there was an issue with the chrono but I fired seven rounds of 9mm factory ammo to check it and they were within 7 fps of each other. Lastly, the Hornady manual shows this load should do about 975 fps through a 7.5" Blackhawk. How should a 5.5" barrel compare?

Thanks

Attached Files

Chronograph 45LC Data N3207.3Gr200GrSWC.pdf (151.6 KB, 96 views)

I think the Standard Deviation of the loads are too great to say that the second string is really faster. The averages are only 1 SD apart.
The velocity spreads on your loads are higher than typical but I don't think I've ever chronographed 45 Colt loads. It may be because the case is so large for the powder charge. Try elevating the muzzle between shots to get the powder at the bottom of the case. Might be interesting to see if it would make a difference.

__________________
Dick

As I understand it (somebody 'splain it to me if I am wrong) To find the range of probable velocity look at the normal SD & avg. vel.

About 65% of the shots will be within the SD +/- from average,

SD x 2 will encompass about 95% of the shots and SD x 3 added/ subtracted from the average will represent 99% of your velocity range.
So...
Test 1 has avg velocity of 712fps and a SD of 25 ...
3x 25= 76 fps +/- the average.

99% of shots will probably be between 636fps to 788 fps

Add all 20 together and run the stats again to get a better picture as 10 is often considered not enough for minimum analysis

You need 10 shots min for avg; 20 shots min for meaningful SD. You're trying to analyze sampling noise; difference is not significant (as defined in statistics) enough to be reliable. The chrono will compute a SD for a few shots, but that does not make it real.

Using a chrono and seeing what you are really getting in your gun with your loads on a specific day can be a shock the first time. Yes, your loads are not consistent, and powder positioning can be a factor unless the case is 90%+ full.

__________________
Science plus Art

How did you measure your powder? Was all the brass the same?

I have chronographed 45 Colt and not had those extreme spreads/SD.Different powder and bullets however.

M1 Chrony? is that the metric version? I had very poor results using a (2) Chrony Alpha Master, very inconstant readings.
(just saying)

__________________
Still Running Against the Wind

Thanks for the input folks. Rule 3, it's a master chrony with the readout in fps. I am not cetain on the accuracy of the chrony but the factory 9mm spread sure read closer than mine. The cases were all trimmed to spec. I'm using a uniflow powder measure with a small cylinder and a digital 750 scale. I feel confident the charges were correct, the small cylinder drop is very consistent. I set the measure and checked that it repeated several times and then every tenth charge thereafter.

I have read somewhere about the volume of powder in the cases being an issue as was mentioned here. Could be. I may try some trail boss or something similiar at some point in the future.

Overall, I am very pleased with the accuracy of the this load but not with the fps results. But, it was fun.

Another factor not everyone looks at is exact case length consistency. If case length varies much, the roll crimp will be different for each cartridge. That will affect the difference in velocities. Now there is absolute exactness or practical exactness. You will have to determine what your standard is. A Bench-rest shooter would worry about .001" in case length or OAL. An action shooter isn't concerned about .025" differentiation as it doesn't affect his practical needs.

__________________
H Richard
SWCA1967 SWHF244

Although I'm also guilty of occasionally using fast powders in large cases, IMO (and you will get debate on it) N-320 is a bit fast for the large capacity case and heavy bullet of the 45 colt and you might look at something a little slower on the burn charts.

I am not a real big fan of Unique but it and HP 38 are my go to powders for the 45 Colt.

Here is when I still had my Chrony and was shooting some of Skip's (Smith Crazy AKA Skip Sackets) hand cast bullets. 250 gr lead.

__________________
Still Running Against the Wind

Rule 3, the spread on your data is far better than what I recorded. I appreciate you posting it. The N320 is fast burning as you guys mentioned and I think I see a powder change in my future at least for the 45 Colt case. To tell the truth, I am using the N320 by default. I bought some of my loading equipment used and the deal also included about 7.5 lbs of N320. I have used it for several different calibers so far and have been happy. However, this was the first time I have real performance data.

Lucky You, that powder is about $35 a pound in my area and it is one of the best powders I have used.... in 9mm, 40, 45, 38 Super 380 and even did good in 38 Spl.

125JHP, yes lucky indeed. Right time, right place kind of a deal. He was getting out of loading and I was getting in. I am using it for everything I load thus far namely 45, 38 special, 357 and 45 Colt. The seller was using it for mainly 9mm using 147 gr PD's for IDPA loads. He explained the power factor etc at the time and most of it was lost on me, still is in some ways. But, it's becoming more clear.

Gamer,

A few statistical observations about your data:

• An F-test shows that your two standard deviations are not statistically different (P=0.18).
• A T-test shows that your two averages are different at the >95% confidence level, even with the large standard deviations.
• A Mann-Whitney W test shows that your two medians (720.25,741.9) are significantly different at the >95% confidence level.
• Even with the high data spreads, you can detect a difference between your two batches. Something in your loading (or testing) process has likely caused this.
• If you require more statistical significance than this, you need to take more data, or find and reduce the cause of variability in the data.

Regards,

Buck.

Wow... did you hear the shock wave as that info passed over my head ?? I can see there is a use to analyze my reloads but I need more info to understand how to use the above tests in a practical way and interpret them. I tried reading the Wikipedia pages on F-test, & T-test and they're way over my pay grade. I envision creating a spreadsheet to analyze chrono & target data to help highlight the best & most accurate loads (at least to my ability)

I know I'm asking a lot and it will be a lot of work, but would you be so kind as to explain how to use these tests to analyze our chrono data and what data elements to use, what formulas, what results to look for and maybe what other various results mean?

or... maybe there is a tutorial for shooters in layman's terms that you could point me to. After looking at Wikipedia I know I need my hand held on this one.

Thanks in advance
Al

Al,

Unfortunately, there is really no short cut for learning statistics. That's one of the reasons I just gave the results rather than the details. If one looks at a batch of data, one is generally concerned with a description of its location, such as mean, median, or mode, and its variability, like variance, standard deviation, or range. Location and variability give you an idea of where the data are and how much scatter they have. A good chronograph will usually calculate mean, standard deviation, and extreme spread which you can use to build a history of your loads. Most people shoot (pun intended) for a low standard deviation as a predictor of good intrinsic accuracy. That may or may not correlate to group size.

The problem occurs when one wants to statistically compare two (or more) different batches of data. It then becomes necessary to employ tests (T-test, F-test) to determine if means and standard deviations are different. The formulas for these tests are straightforward but tedious. That's why most folks use computer software for the calculations. Microsoft Excel has routines that will calculate the basic statistical tests.

However, one still has to interpret the results. You may want to look for an elementary text (like "Statistics for Dummies", if that exists) to study. Like I said, no shortcuts - you just have to bite the bullet and learn.

FWIW, I've spent 50 years studying this stuff, and I'm still learning.

Sorry I couldn't be more help,

Buck

What forum is this??This is getting to deep, sounds like work

I thought the primary factors were if it went bang and I hit the target? I refuse to buy the cable adapter for my Competition Electronics Pro, as I am not bringing a LAPTOP to the range.

__________________
Still Running Against the Wind

I don't take a laptop to the range either but later in the day, when I'm looking at the results, I always seem to end up crunching numbers to make comparisons between loads or a load shot from different guns. Buck turned on a light bulb in a dark room that might help with that process.

Thanks for the reply Buck, I'm not really looking to learn all about statistics, just some specific routines to apply to chrono and target results to help indicate if a load is valid, good, bad, erratic or whatnot when comparing it to another similar one. Sorry the following is a bit wordy, I get carried away easily...

I currently use a CED M2 chrono to collect each shot velocity and then use Excel to calc the Avg.fps, SD, MAD and ES. Of those, I compare the Avg.fps to the expected calculated from the load manual or OEM published fps and also look at the SD & ES for low numbers, but not knowing what is "commonly accepted as low", I had to arbitrarily come up with my own standards (if that is what you would call them). I decided to consider a SD less than 20 as good and one less than 10 as Excellent. As far as Extreme Spread, I have allowed a little more range such as 50 fps or less as OK, 30 as good, 20 = very good and 10 as excellent. It's things like this I get cloudy on. I also recently started calculating the 95% & 99% SD in order to evaluate what the true velocity range might be with the chrono data I am seeing.

I scan my targets with a 1/10" grid overlay to .jpg files and then use "On-Target" to determine avg. group size and ES and POI offset from POA. I then enter the load data and target shot info into "Point Blank" software for archiving records and to try and compare which bullet/powder load is the better, in a particular gun. Pretty much what everyone does but a little more anal in some ways I guess. I can see where comparing data using the tests you described can help identify good/poor loads when it isn't as obvious.

I can usually figure out a lot on my own, and am pretty good with Excel. I think I was overwhelmed with the unfamiliar terminology presented in the Wiki pages and how much I had to drill down to figure out what several words meant that they were using. I realize terminology like "mode" and "location" etc. have specific meanings to statisticians, and I may pickup some of it as I learn but I don't want to know it intimately, just enough to get the job done and layman's terms would be better to understand it all.

I found the F-test & T-test functions in Excel - now I have to learn how to use them. Any other Excel functions you can suggest I learn to use?

Rather than a tutorial, perhaps you or someone could provide a little help to get started in the right direction. A couple specific questions I have are,,,

When would one choose to use the F, T or W test and what would each try to indicate?
Why use one over the other?
What do I compare?
What data do I need or want to calculate?
What are valid / invalid ranges to look for?

Just from the little bit of info in the above post I concluded the following....
o- You used the F-test to compare the SD of two strings to see if they are different, which you conclude they are not. This implies to me that they are within a comparable scope or are similar enough to each other to continue the comparison. correct?
- what is considered 'different'? what value range would be good enough comparing 2 or 3 SD's to say they are the same for our purposes?

o- You used a T-test to compare averages of the shot strings and concluded they are far enough apart at the 95% level to be different. This seems to imply there is a problem relating the two data sets as similar or ~equals. Again, how would one tell what results or ranges are valid

o- When talking about the W-test, you use the 'medians' and provide median numbers, did you use a value from midway in the string or calculate the middle-value from the range?

In summary, am I correct to assume that you initially compared the SD's with the F-test to quickly see if they were considered different (as opposed to similar), then you processed the average & median data (of each string) out to its 95% extreme (based on SDx2) and made comparisons with the T & W tests to see if they were close to each other (being similar) (from 2 different perspectives-avg.&median) and found that they were too far apart to be the considered from similar data.

This then indicated that even though some data points overlap in a common expected range, there is something else causing them to act as different loads altogether. Your suggestion is then to look at the load process for the variable.

One of the things i like about these forums is the chance to learn new things. You have opened up an interesting avenue I will try to explore further.

Thanks,
Al

Gamer,

Suggest the folowing to check your ammo for powder position sensitivity.
(In other words, start checking for possible reasons the loads vary. I teach statistics, and I assure you further analysis of your present results will only confirm you have a problem, NOT fix it.)

Fire two strings of 15 shots each, and record results for each.
First string, arrange the chronograph so the gun points slightly UP when you shoot. Point the gun straight up, then slowly bring it down and shoot.

String two, arrange the chrono so the gun points slightly DOWN. Start by pointing the gun straight down, and then bring it slowly back up to fire.

I think you'll see a large difference between the means of the 2 groups, and the SD for each group should be less than previous.

If this is true, as I suspect, it means your powder does not fill the case enough for you to get uniform velocities unless you preposition the powder each shot. Solution: change to a powder with higher loading density; this is why they made trailboss for those big pistol cartridges used in Cowboy Action.

__________________
Science plus Art

Al,

I'll try to get a detailed answer done for you tomorrow (it's a bit late now). I think you're on the right track, and I'll answer your specific questions then.

Buck

Al,

I've added comments below.

Extreme Spread is not usually a very good estimate of variability, since a single point can skew the estimate to extremes. Variance, Standard Deviation, or Median Absolute Deviation (MAD) are much better choices.

This is an excellent thing to do if you're serious about studying loads and accuracy. In particular, comparison of average group size with velocity standard deviation will show you how good the assumption is that lower SD leads to smaller groups. The assumption (in my experience) fails more often than you would expect. But it almost never fails in the direction of low SD leading to larger group sizes.

One important reason for checking the equality of standard deviations is the T-test calculations (for deciding about the difference in means) are dependent upon whether the standard deviations are equal or not. In Excel you will see two different T-tests for the two possibilities.

Obviously, another reason for checking standard deviations of two groups is to see which has the larger velocity spread, and if that is a real difference.

The one thing you have to do with these statistical tests is to decide how certain you want to be about them. Most people pick 95% (1 chance out of 20 the test is wrong). The value you choose depends on how important the decision is. A 95% chance of being right is probably OK for something like reloading. I suspect that you would like the chance of the airplane upon which you are flying not to crash to be considerably higher than that (like 99.999999999...%). You pick.

The median is the middle point in the batch. For an odd number of points, that is the actual middle point. For an even number of points, it's the average of the two central points. The median is a very robust estimate of the location of a group, and can't be skewed by a single point (large or small) as easily as the arithmetic mean can, especially for a small number of points. The disadvantage of the median is that it is a less powerful test than the test for means. That means that it takes more data for the median W-test to draw the same conclusion about equality of location as the mean T-test. So why do we use it? The median test does not depend on the data being Normally (bell-shaped) distributed. If you have a highly non-Normal data set, the F- and T-tests can give you the wrong answer. So I check both.

No, you include all the data. The T-test computes the difference between means and compares that to a statistic dependent on the variances of the two groups, the % confidence you require, and the number of data points in each group. If the difference between means is larger than the statistic, the difference is statistically significant.

Regards,

Buck

Is it too late for drop/add for this course??

Man, I am a chemist and physicist by education, but I really never liked getting too deep into the statistics, although in some aspects of those disciplines we are all about statistical probabilities.

The chrono results look pretty typical for what I obtain with mixed brass handloads of indeterminate case length. I attribute that to be an issue of consistent crimp as much as anything else.

The powder you selected is a good one. Is it optimal for charge and bullet weight, i.e. does it do better with light for caliber bullets or heavy ones, or is it at near max pressure to be within its best performance envelope, how is load density and is it position sensitive?

I do not particularly like Unique as I find more velocity variation with it since it is among the more difficult to meter consistently, imho. I have developed 45 Colt loads using WSF and find the ES and SD values, as well as accuracy, to be pretty good with it with its optimal loads, metering being only one of the factors I'm guessing. Hitting the sweet spot is a matter of trial and error, but a good chrono helps. Still, sometimes your best chrono statistics don't result in the most accurate load. Go figure.

[Note: There is a 100% probability I will have spelling or syntax errors requiring correction in nearly every post I make. The correlation coefficient of that regression anaylsis is 1.00]

__________________
" I said, good DAY! "

Thank YOU very much Buck, I will read that several times as there is a ton if info in there I need to completely digest.

As a learning example, I transferred the data from the pdf to an excel spreadsheet. I successfully ran the Excel F-test and T-test functions and came up with the same results as you mentioned. I also found a couple of ways online to do the W-test (aka U-test) formula. and a simple explanation of its use. ( Excel Master Series Blog: The Mann-Whitney U Test Done in Excel )


o- I discovered that using the median function can return the same value as the average function, or a different value, depending what is used as input.
1. I calculated the average of the 2 10-shot chrono strings (712.2 fps & 747.9 fps).
2. I then ran the median function on the 10 shot chrono strings and received different values (720.25 & 741.9) which correspond to your results.
3. I also calculated the 95% & 99% hi/low values using the average fps and the SD x2 & x3. When I ran the median function using these min/max values, the result was exactly the same as the average of the raw chrono strings.

o- I ran the F-test function using the two 10 shot strings as input and received a value of .18

o- I ran the T-test function using the two 10 shot strings and picked a 2 tail distribution and selected type= 1 (paired) - I have no idea (yet) what they mean but I had to pick something. the value returned was .027

I think I am making some progress in understanding all this and will keep reading on the subject. Your suggestion for a 'dummies' book may just be what I need (in addition to a nap after all this mental exercise) as like you said, there isn't any easy shortcut to understanding it and I see now that I will have to understand the terminology to understand how to get the results I want.

Thanks again
Al

Al,

The two-tailed distribution is the correct one to use (by most statisticians for most problems), but the data are not paired - shot #1 of one string has no correspondence to shot #1 of the other, and so forth. However, P=0.027 was about what I got with unpaired data analysis, so it must not matter much for these sets. Paired data would be something like comparing two different types of shoe soles for wear. Sole A would be on the left foot and sole B on the right for a single individual. All other individuals would wear the shoes the same way. That takes any individuals' differences (height, weight, etc.) out of the analysis, which is the reason the trial is done that way.

The P-value is used to assess the truth of the null hypothesis (the means are equal) versus the alternate hypothesis (the means are unequal). If the P-value is less than the significance level of the test, the null hypothesis is rejected and the alternative hypothesis is confirmed. In this test, the significance level a is 0.05 since (1-a)=0.95 (95% confidence level). Since (P=0.027) is less than (a=0.05), the null hypothesis is rejected, and the means are not equal. Note that if you had wanted 99% confidence that the means were different, the null hypothesis would have been confirmed.

The same interpretation of P is used with the F-test, except you are comparing variances (or standard deviations). Null = equal variances, Alternate = different variances.

Buck

Thanks for those details Buck, they clarify a number of other questions I had and the examples helped me understand the process considerably. The process is starting to make sense... or as I might say in my kind of terminology.... the mud is settling in the pond, I can see the fish.

I changed the T-test type to '3' (two-sample unequal variance) which now sounds to me to be more appropriate than a paired test.

a couple of things I have learned from this...
For reload testing, 99% is probably unnecessary, I can just evaluate at 95% levels.
I need to start thinking in terms of means and variances rather than averages and ES.
Hopefilly good interpetation results will follow.

Thanks for taking the time to explain all of this,
A question I probably should have asked in the beginning... The OP was comparing two sets of the same load fired from the same gun.
Is this process and these same tests, applicable to comparing loads with different powder charges, (ie: Same bullet, Similar velocity, different powder & charge... to determine which is better to load up 5000 of) ... or... to compare the results of the same load shot from different guns?

Al
PS attached is a XLS for readers info.

Attached Files

Stats.zip (13.0 KB, 2 views)