Range Report Question about ES and SD for smart types

[Steel head]

I had a 10 shot string with ES of 21 and SD of 6.

Moved to another location and had an 11 shot string with ES of 20 and SD of 6.

Had that been one continuous string would the SD still be 6ish?

 

[J McGriddle]

I think would that depend on the ES for all 21 shots.

 

[Huskydriver]

Depends on where the average velocity lies for both shot groups...es and sd are measurements from an average point...SD of 6 just means that 68% of your shots fell +- 6fps of the average velocity given you have a normal bell curve sample.....

Es just means that the diff between the farthest outliers was 20 fps.


Its possible that the average for one group was 2700 fps and the other was 2800....and you could still get an SD of 6 and an es of 20

 

[Howland]

If the average of each group is the same or very close, SD will probably be very close to 6.

Do you have access to Excel? You can plug in all 20 velocities and run a formula.

Define "Moved to another location." Same day, altitude and atmospherics won't have an impact. The chrono being several feet closer or further could make a slight difference which could show higher SD that really isn't there.

ES of 20 or 21 with SD of 6 over two, ten shot groups is pretty good. Group sizes?


I think you're just bragging. I'd be pretty proud of myself too.

 

[Steel head]

Depends on where the average velocity lies for both shot groups...es and sd are measurements from an average point...SD of 6 just means that 68% of your shots fell +- 6fps of the average velocity given you have a normal bell curve sample.....

Es just means that the diff between the farthest outliers was 20 fps.


Its possible that the average for one group was 2700 fps and the other was 2800....and you could still get an SD of 6 and an es of 20

Yea
I should have taken down all the numbers.
I can say all 21 shots had a low of 2874 and a high of 2895

 

[Huskydriver]

Yea
I should have taken down all the numbers.
I can say all 21 shots had a low of 2874 and a high of 2895

I would say your safe bro .... Have you ever ran the diff for 35 fps in dope at 1000k in 6.5 Creed?

 

[Howland]

So ES of 21 for twenty shots means your averages are almost identical so your SD should shrink slightly (about 4.5%).

 

[Steel head]

If the average of each group is the same or very close, SD will probably be very close to 6.

Do you have access to Excel? You can plug in all 20 velocities and run a formula.

Define "Moved to another location." Same day, altitude and atmospherics won't have an impact. The chrono being several feet closer or further could make a slight difference which could show higher SD that really isn't there.

ES of 20 or 21 with SD of 6 over two, ten shot groups is pretty good. Group sizes?


I think you're just bragging. I'd be pretty proud of myself too.

Lots of people get better SD’s than me but I do ok for being minimalist(lazy) and coincidently that’s the exact SD my last barrel consistently did.

I shoot the same steel and rocks from multiple locations.
It was about 5 minutes of driving to the other location.
I mainly recorded data today because I’m upgrading my scale soon and want to see if there’s an improvement.
I almost never shoot groups but I did shoot 5 shots on paper today as I changed out my rings for higher ones and had to re zero.
Here’s all 5 shots.

 

[Steel head]

I would say your safe bro .... Have you ever ran the diff for 35 fps in dope at 1000k in 6.5 Creed?

For 6.5 CM no because

I have for my 260 though

 

[Nik H]

Depends on where the average velocity lies for both shot groups...es and sd are measurements from an average point...SD of 6 just means that 68% of your shots fell +- 6fps of the average velocity given you have a normal bell curve sample.....

Es just means that the diff between the farthest outliers was 20 fps.


Its possible that the average for one group was 2700 fps and the other was 2800....and you could still get an SD of 6 and an es of 20

This right here

 

[lash]

It’s already been covered pretty well, but to summarize; you need to run all 21 shots as a string in excel or similar to get the true SD for all shots. Same for ES.

But I don’t think that you have much to worry about. Those are good solid numbers. It shows that you have good consistency in our loading practice.

 

[Ledzep]

I'm sure you've read my opinion on ES/SD/Group size before...

If you are using a sample size of under 30-35 your confidence level in any of the statistical analysis tools is lower than most people would use to make serious decisions.

When people post SD's of much less than about 7 or 8, you can call BS most of the time. Not every time. But most of the time. If you ran those same 5-shot samples out for 35-50 shots you'd watch the SD climb to 8-13fps. Same goes for ES much below 25fps. Again, I'm not saying it's impossible, just that it's very EASY to get a small SD in 5 shots, and very HARD to keep it small for the long run. The long run is what matters when you shoot 90-250 rounds in a match. Most people want good numbers instead of facing reality.

 

[lash]

I'm sure you've read my opinion on ES/SD/Group size before...

If you are using a sample size of under 30-35 your confidence level in any of the statistical analysis tools is lower than most people would use to make serious decisions.

When people post SD's of much less than about 7 or 8, you can call BS most of the time. Not every time. But most of the time. If you ran those same 5-shot samples out for 35-50 shots you'd watch the SD climb to 8-13fps. Same goes for ES much below 25fps. Again, I'm not saying it's impossible, just that it's very EASY to get a small SD in 5 shots, and very HARD to keep it small for the long run. The long run is what matters when you shoot 90-250 rounds in a match. Most people want good numbers instead of facing reality.

You are correct of course, but those were 10 and 11 shot strings, not 5 shot strings. Same principle basically applies, but there is higher probability of statistical confidence in his 21 shots. And for what most of us do here, those are plenty good numbers to move forward. If you think I’m going to spend the time and barrel life to waste 30 round strings on everything so that I am 100% secure in my probability analysis, for typical field match and PRS style shooting, you are mistaken. Neither needed nor necessary. Especially when using a known load in a known cartridge.

 

[Ledzep]

My comment about 5 shot groups was more in response to Steel Head saying that others get better SD's than him. 21 shots with that spread is enough to say that he's onto something.

I still maintain if you're not going to run a large enough sample size to make the tool useful, there's no point in recording or posting it. I've done enough large sample size testing, and continue to do more and more of it and the more I do, the less work I do for my own load development because it's showing me that most of the "patterns" I used to chase were/are noise. eta: *Que everyone telling me my test parameters are shit and I don't know how good THEIR stuff shoots*

 

[Steel head]

My comment about 5 shot groups was more in response to Steel Head saying that others get better SD's than him. 21 shots with that spread is enough to say that he's onto something.

I still maintain if you're not going to run a large enough sample size to make the tool useful, there's no point in recording or posting it. I've done enough large sample size testing, and continue to do more and more of it and the more I do, the less work I do for my own load development because it's showing me that most of the "patterns" I used to chase were/are noise. eta: *Que everyone telling me my test parameters are shit and I don't know how good THEIR stuff shoots*

I shot those strings today on steel and rocks from 1787 yards to 2997 yards.
Vertical was excellent just like that load was in my last barrel.
I do agree that more shots gives a better sample but I find these loads with a two shot per step speed test then confirm Prospect with a 10 shot string at distance.

 

[lash]

Edit: this is a response to Ledzep’s last post.

I agree. But, as you yourself say and many of us having done it a while know, there are ways to assure that you are in the right zone. When you already have experience with a cartridge, throwing rounds down range to verify same becomes redundant and more an exercise in self assurance than a productive exercise.

 

[Dthomas3523]

This is why I just focus on my ES. It’s easier for me to wrap my head around when using chrono and not recording all shot strings in spreadsheets.

I can easily remember the highest and lowest shots I’ve seen on a chrono every time I run it the week before a match.

So, if I know for example my low is 2940 and my high is 2960. Then I chrono this week and it’s 2935 and 2975, now I just remember those numbers. Or write them down.

And I don’t care what my sd is because while you can have a low SD and a high ES, you can’t have a low ES and a high SD.

My ES threshold is .1 up or down @ 1k. I take my trued data and run numbers on what the furthest point from my average each way that will move more than .1 up or .1 down. If I get any numbers higher or lower, then I know to take a look at what’s going on.

For example, I’m running 109’s @ 2950. Or about 7.4mil@1k. I can go as low as 2920 for 7.3 and as high as 2965 for 7.5.

So whenever I chrono, usually the week before a match or any range trip......I just briefly look at the sd and make sure it’s acceptable (it always is when you are in a node and using good loading technique/equipment), and the low and high is inside 2920 and 2965.

 

[Milo 2.5]

I shot those strings today on steel and rocks from 1787 yards to 2997 yards.
Vertical was excellent just like that load was in my last barrel.
I do agree that more shots gives a better sample but I find these loads with a two shot per step speed test then confirm Prospect with a 10 shot string at distance.

I have been using a V3 for about a month, on 2 existing loads, I have not seen a improvement in numbers or downrange accuracy. One thing I did not do, and I should have was verified the chargemaster load against the FX 120i. I may be throwing over or under the CM average.
Now I also got a new 6BRA at around the same time, so all charges thrown have been V3. I was shooting in 20+ mph winds yesterday, and the load held vertical at 1K. I chrono'd my last 15 shots, and they were not as good as I expected. Regardless of numbers, the load has to hold vertical at the distances you choose to shoot at. I have 3 test samples of varying neck tension that against ledzeps advice, I am going to group at 500, take the best, and I am positive my numbers will shrink with the best group, and I should be on my way to the range instead of here, nice morning.
IMO, true test of any load that is numbers based is to shoot it at first light, then again at around 11 am, then 4 pm. Your total ES may expand, but if SD remains stable within the 3 temp ranges, good to go.

 

[Milo 2.5]

I'm sure you've read my opinion on ES/SD/Group size before...

Did every educator at SDSMT have the exact same philosophy?
I'd bet not, or every suspension bridge made would look the same.

 

[Ledzep]

Same load, same gun, 3 separate tests. This is velocity data. Shots are in order along the bottom, along the left is the running SD in FPS (i.e. at 2 shots is the first data point, at 10 that's the SD for those 10 cumulative velocities up to that point, etc..). I was testing new cases vs. once fired, vs. once fired stripped clean. New cases in the middle below for those curious.

0-10 shots is noise. About 15-20 you start getting stable velocity data. I only shoot 30-35 for these tests anymore. There are other metrics that still have more variance beyond 15-20 (Dispersion, mean radius and radius SD mainly). This cost you nothing. Interpret your own conclusions.

 

[Milo 2.5]

Same load, same gun, 3 separate tests. This is velocity data. Shots are in order along the bottom, along the left is the running SD in FPS (i.e. at 2 shots is the first data point, at 10 that's the SD for those 10 cumulative velocities up to that point, etc..). I was testing new cases vs. once fired, vs. once fired stripped clean.

0-10 shots is noise. About 15-20 you start getting stable velocity data. I only shoot 30-35 for these tests anymore. There are other metrics that still have more variance beyond 15-20 (Dispersion, mean radius and radius SD mainly). This cost you nothing. Interpret your own conclusions.

I hope it doesn't seem like I am on your ass constantly, you know I kinda like you, lol
One of the biggest travesties since the inception of the PRS is the fact that all us so called experts with our ways of doing business has caused the 1K benchrest guys to clam up. I don't give a shit who you are, we all want a 1" shot group at 1K, just once. Maybe today that group is at 2K, or farther.
There is one guy, maybe more, that talk of tuning a load periodically throughout the day to win. Numbers change over the coarse of a day, even shooting a 35 shot string one can physically enter variables into it, fundamentals, hot chambers, etc....
In my world, the more I fire at a target, the more I collect numbers data, the faster it sours. Though the numbers aspect may be desired.
Lol, I have always said this, concerning # of shots needed for group size, no one has ever shrunk group size after 3 shots, realistic or not.

 

[Steel head]

I hope it doesn't seem like I am on your ass constantly, you know I kinda like you, lol

 

[Milo 2.5]

Same load, same gun, 3 separate tests. This is velocity data. Shots are in order along the bottom, along the left is the running SD in FPS (i.e. at 2 shots is the first data point, at 10 that's the SD for those 10 cumulative velocities up to that point, etc..). I was testing new cases vs. once fired, vs. once fired stripped clean. New cases in the middle below for those curious.

0-10 shots is noise. About 15-20 you start getting stable velocity data. I only shoot 30-35 for these tests anymore. There are other metrics that still have more variance beyond 15-20 (Dispersion, mean radius and radius SD mainly). This cost you nothing. Interpret your own conclusions.

I hate typing lengthy responses, but if you are gathering numbers that you feel is solely important, are you culling brass based off case capacity? Are you culling loaded rds based off neck tensions or seating force?
My God, we buy AMP annealers thinking we get a 100% uniformity in annealing, when the brass we run in it does not have a 100% equal composition making it impossible to do. Then we seat bullets in that brass thinking again all is equal.
I understand the numbers in science, and sample size, but unless we remove every friggin variable we can upfront, results are most likely skewed before we start.

 

[lash]

Same load, same gun, 3 separate tests. This is velocity data. Shots are in order along the bottom, along the left is the running SD in FPS (i.e. at 2 shots is the first data point, at 10 that's the SD for those 10 cumulative velocities up to that point, etc..). I was testing new cases vs. once fired, vs. once fired stripped clean. New cases in the middle below for those curious.

0-10 shots is noise. About 15-20 you start getting stable velocity data. I only shoot 30-35 for these tests anymore. There are other metrics that still have more variance beyond 15-20 (Dispersion, mean radius and radius SD mainly). This cost you nothing. Interpret your own conclusions.

View attachment 7395064
View attachment 7395065
View attachment 7395067

I also want to go on record that I appreciate the work that you do and that you are sharing with those that care. Having said that, I took a good look at each of three graphs above and to my eye (having used practical analysis my whole career), there is little practical difference between the SD numbers at 10 shots and those at 30 shots in all three cases. Without actually knowing the data points themselves and merely looking at where on the SD scale the curve is at both 10 points and 30 points, it seems that the SD at 10 data points is easily within 10% of the SD at 30 data points.

What exactly am I to gain by tracking those extra 20 data points anyway? It certainly seems that the SD has mostly stabilized by the time it hits 10 data points, at least this is what your small set of three charts seems to indicate.

Interestingly, coming from the automotive OEM arena, I completely understand that 30 data points is the minimum statistically acceptable number of points need if one is expected to be able to show stability and a true SD. Typically 100 data points are required to show process capability.

What we should keep in mind however, is that a process that has been shown to be stable/in control can yield useable results using fewer data points. This is especially useful when making incremental changes or improvements to a process. It allows one to test multiple incremental changes and narrow down to those that are improvements versus those that are not.

 

[Ledzep]

I've got a new rig fixed up that eliminates shooter error so for the time being I'm repeating previous shoulder-fired stuff to sanity check. Anyway, I'm working through one variable at a time, starting with the basics and moving into the more asinine stuff as I go. Maintaining the "good" practices as I go to hopefully reduce noise along the way. It's going to take me a year or more just with what I have slated. End of the year goal is to have definitive say-so in what makes what differences.... in a then shot-out barrel ... But seriously I tend to believe the trends I'm looking for will carryover from barrel to barrel, and from what I've already done I have seen that play out (so far).

Ultimately you have to put into perspective what it is you want to accomplish. If you want to shoot 1-2 MOA steel plates at 400--600yd you might as well not even buy a chronograph. If you want to shoot the same targets (1-2moa) at 800-1000yd you probably want to monitor extreme spreads a little. If you want to hit 1-2 MOA targets past 1000yd you need to know a very accurate average, and have reasonably tight ES. If you care about your first shot hitting (hunting), you need accurate average, ES, SD, and apply 4 or 6 sigma to hit probability calculators with trued ballistic software (Just to find that you probably shouldn't shoot at animals past 600-700yd)... That is the real purpose of a SD, and it requires your data to have a roughly normal distribution to be accurate... And if you're not doing that, there's really no point of keeping track of it. That's all I'm saying. It's a little lie you tell yourself to feel good when you shoot 5-10 shots and have an SD of 3fps. It's a statistically insignificant number with so much chance for variance that it's pointless. You could just as easily have got 14fps SD and thrown the EXACT SAME LOAD out the window.

If you want to shoot 1" groups at 1000yd, buy a lottery ticket. There's a reason the world record 10-shot group at 1000 is 3.5x larger than the world record 5 shot group. It's not because the guy shooting 5 shot groups has a better rifle. And don't get me wrong, the BR guys are some very talented shooters and reloaders. I'm most interested in practical and reliable field use and understanding limitations and hit probability at UKD targets in varying terrain. If I have to tune the load the day of, it does me little good.

 

[Ledzep]

To better answer your question, Milo, I'm recording bullet seating runout and will eventually have access to a load cell that will let me plot data vs. seating force.

 

[steve podleski]

Google says:

after we square the SDs, we divide each by its own sample size. Then add square root to get the combined standard deviation.

 

[Huskydriver]

Google says:

after we square the SDs, we divide each by its own sample size. Then add square root to get the combined standard deviation.