If you want a good look at the effect of sample size, this video makes for an interesting discussion.
Hornady Method: Revolver 45 Colt
- Thread starter Nathan
- Start date
stagpanther
New member
Ha--that's pretty good--like seeing the forest through the trees.
Yep. That's what statistics is supposed to help us do. The problem for the human brain is it is designed to search for and find patterns in things, which it is good at and can do even with random data. The old TV cop adage "I don't believe in coincidence" is an excellent example. The late Isaac Asimov pointed out long ago that the world would be a much odder place if there were no coincidences. Nobody could ever win a lottery because a ticket number matching a picked number is a coincidence. No good or bad luck would be possible at all because those are coincidences of events and changing circumstances. Nobody could meet their soul mate.
There's a really good 2014 program on NPR's RadioLab show called Stochasticity, which does an unusually good job of explaining and illustrating how easy it is to see a pattern that's not really a pattern at all or to think something is not random when it actually is. I highly recommend giving it a listen, even if you aren't normally an NPR fan. Always good to have a little healthy skepticism about what we think we are seeing. That makes us try to prove it one way or another.
There's a really good 2014 program on NPR's RadioLab show called Stochasticity, which does an unusually good job of explaining and illustrating how easy it is to see a pattern that's not really a pattern at all or to think something is not random when it actually is. I highly recommend giving it a listen, even if you aren't normally an NPR fan. Always good to have a little healthy skepticism about what we think we are seeing. That makes us try to prove it one way or another.
stagpanther
New member
The random generated stuff reminded me of the first Litz books; that was heavy slogging through the trenches getting through them for me. Funny thing is, in my younger days I had an MBA in project management which majored in statistics--I was really good at it but over the years the artistic side of what few brain cells I have left got the better of me (maybe shooting report concussion as well)--and I forgot almost everything. I only remember the 1st commandment from my very first stat 101 class: "Statistics don't prove anything--they only suggest correlation for decision-making."
As for the "meaning in randomness" thing, that probably evolved as a survival mechanism--the word stochastic is one I haven't heard in a long time; I also did a brief stint in neurophysiology and the the nature of the eye's "micro-saccade" as a stochastic input to the brain and served as a survival mechanism was fascinating to me.
PS--funny enough the NPR thing you suggested has a balloon as an example--coincidence?
As for the "meaning in randomness" thing, that probably evolved as a survival mechanism--the word stochastic is one I haven't heard in a long time; I also did a brief stint in neurophysiology and the the nature of the eye's "micro-saccade" as a stochastic input to the brain and served as a survival mechanism was fascinating to me.
PS--funny enough the NPR thing you suggested has a balloon as an example--coincidence?
Last edited:
Once you generate your “data” by random number generation, you are no longer talking about shooting. You are talking about number generation.
Shooting does not generate velocities randomly. It is a multi variable process that we are recording velocity data from. The reason each shot differs could be dimensional differences, bullet imbalance, burn rate, bullet weight, charge weight, primer sensitivity, etc…all variables are different each shot, yet we measure velocity and group size to determine what happened!
Shooting does not generate velocities randomly. It is a multi variable process that we are recording velocity data from. The reason each shot differs could be dimensional differences, bullet imbalance, burn rate, bullet weight, charge weight, primer sensitivity, etc…all variables are different each shot, yet we measure velocity and group size to determine what happened!
Recycled bullet
New member
"...multi variable process that we are recording velocity data from. The reason each shot differs could be dimensional differences, bullet imbalance, burn rate, bullet weight, charge weight, primer sensitivity, etc…all variables..."
What about relative caffeination, blood pressure and heart pulse rate, eye sight acuity, steadiness and strength in the fingers, the guy in the next lane mag dumping an AR10 SBR?
The gun variable is just one facet of the shooting skills diamond.
What about relative caffeination, blood pressure and heart pulse rate, eye sight acuity, steadiness and strength in the fingers, the guy in the next lane mag dumping an AR10 SBR?
The gun variable is just one facet of the shooting skills diamond.
Recycled bullet
New member
I like to do stretches and inclined pushups then while my heart is up shoot pistols off hand for groups at 30-50 yards at nine inch paper plates.
If I can teach myself to shoot competently under less than ideal conditions I am conditioning myself to shoot competently under all conditions.
If I can teach myself to shoot competently under less than ideal conditions I am conditioning myself to shoot competently under all conditions.
Nathan said:
The velocities have a mean value controlled by the gun and the load recipe values. But the variations in velocity around that mean are still randomly distributed and follow the gaussian normal distribution. This is why your chronograph includes a standard deviation calculation. That calculation assumes a randomized distribution of variance that is normal.
Nathan said:
That's still a collection of variables that, individually, have normal distributions of magnitude and location and that are randomly distributed by the time they get into the gun. We produce rounds we know have some variation but don't necessarily succeed in identifying and tracking them in the ammo box. So they get randomized in the same way mixing a hat full of numbers for a drawing randomizes them. What you will find is they produce results that generally fit the normal random distribution. Remember, it's the distribution of these variables that is random and not necessarily the individual causes of variation that are random. It's a fine distinction but a real one in that it enables statistical manufacturing controls to work because we can identify and neutralize causes of variation.
For combat simulation, I recall someone forty years ago making the suggestion that you bench the weapon and run as fast as you can back and forth behind the firing line until your heart is racing, your palms are sweaty, and your forehead sweat is starting to run into your eyes. Then pick up the gun and see what you can hit with it and how quickly. The trick is to get your heart rate to exceed 65% of your maximum heart rate. That's the magic number above which fine motor coordination falls apart, as commonly occurs during an actual fight or when you experience a fight or flight response to a life-threatening situation.
GeauxTide said:
The stat terminology can be confusing, but I think you mean variation. Variance is the square of deviation from the mean. So if you have a ten-shot sample with a population standard deviation of 10, the mean of the variance will be 100 (10²), even though the extreme spread (variation) is only around 30.
That looks better than my 20 shot group!
That's 20 rounds of factory Hornady Black 105 grain BTHP 6mm Creedmoor, from the same lot even. The data wasn't even that great.
I realize I'm comparing apples to orangutans here. When reloading I try to get my standard deviation in the teens for ten rounds. This is the first time I shot a 20 round string other than in rimfire.
The only reason I'm shooting factory ammunition as well is I got 400 rounds of Hornady Black for $400 delivered. It was cheaper than buying the brass!
That's 20 rounds of factory Hornady Black 105 grain BTHP 6mm Creedmoor, from the same lot even. The data wasn't even that great.
I realize I'm comparing apples to orangutans here. When reloading I try to get my standard deviation in the teens for ten rounds. This is the first time I shot a 20 round string other than in rimfire.
The only reason I'm shooting factory ammunition as well is I got 400 rounds of Hornady Black for $400 delivered. It was cheaper than buying the brass!
Small spreads in velocity giving low SD and ES numbers are nice, but what matters is what the gun/ammo does in your hands.
Some guns, particularly the larger bores, are less sensitive to small variations than others. Variations in loads or in barrels that make a significant difference in a small bore might not matter is a large bore, as they are a much smaller percentage of the overall factors.
Some guns, particularly the larger bores, are less sensitive to small variations than others. Variations in loads or in barrels that make a significant difference in a small bore might not matter is a large bore, as they are a much smaller percentage of the overall factors.
Shadow9mm said:
You need to know you are operating in a safe pressure window…before shooting your 20. The 20 is just saying you have enough data points to really know the zero, accuracy and velocity of your load. The method is load 20 of something and shoot. If you get 5 on paper and your group is too big, well, it is not getting smaller, so pull the rest and try again. If you get to 20 without blowing up your goal, then it it saying you can reasonably expect this performance when you are counting on it.
Basically they said when we develop on 5 shot groups that none of the groups is a sweet spot and if we shot 20 of each, they would likely all be quite similar. Miles also suggested that lighter charges are always slightly more accurate in his experience. So, load as close to your minimum acceptable velocity and give it a go. If too big, pull, change and try again.
My point with the am I done question is this load meets my hunting needs and was tuned in 20 shots. Sure, it could be better. I may try a grain less or 2400 powder next time, but I know this will be a reliable 3” 25 yd shooter. I won’t find it shoots 7” groups because my sample size was too small.
stagpanther
New member
It's been almost 40 years since I studied statistics--so I only have random recollections and can pick out only bits and pieces from memory. I hope unclenick will weigh in with his knowledge--he clearly has a sharp command of mathematical and engineering concepts.
My recollection of Litz's studies--was that he used computer-generated simulated shot dispersions with pretty large population fields to show that there was correlation between the simulations and what he shot in the field (that's a rather over-simplified interpretation of mine, Litz wrote two or three volumes on the matter). Sample population field size (how many of what you are studying) appears to be what constantly comes up in these discussions. The confidence interval in correlation doesn't prove anything with certainty, it suggests a degree of probability that can be expected for an outcome. Computer simulations are used to design and simulate tests of the US' nuclear weapons--doesn't seem far-fetched to me that they can be used as a probability tool in simulating shot dispersion.
Last edited:
My point most be missing…..computer simulations are quite common in product development. Each component of the system is simulated with the full or partial laws of physics. Then the system outcome can be viewed.
For example, I can apply a heat source to one end of a bar in simulation and measure how the other end heats up. That gives me a result based on those inputs which simulate the actual system.
If I plug in a range of temps into a random number generator driven solely by my inputs, I will get my input data back as data points, but it is unrelated to the system I’m evaluating. That is my point.
Hornady, and I suspect Litz, use these methods only to show what it should look like….not showing it as actual system results. Hard not to get sucked into the random number generator. It can only be used to illustrate a point.
Nathan said:
That's the disconnect for me. Having done a lot of finite element modeling for my insulation patents, I am very familiar with the process of modeling material properties. But once you've done that, how can the response you get to any kind of source of heat or force or whatever, random or not, be unrelated to how the system would respond to those same conditions? In the modeling, you've gone through a lot of trouble to ensure the response is very specific to the system. So I am missing something there.
The random numbers generated for target hole locations are to represent noise in the shooting system. They are weighted by the probabilities of the Gaussian normal distribution that is ubiquitous to noise throughout nature. You take a thousand men and make a bar graph of their weights, and it forms the bell curve. If you measure the thickness of sheets of paper in a reem, you will find the distribution of the thickness variation forms the bell curve. Like shooting, these bell curves are the result of mechanical processes, but the "noise" in the processes follows the normal distribution. The assumption that this is so for noise in mechanical processes is the basis for six-sigma quality control, which is a tested and tried and true methodology.
For groups from a gun, in particular, describing them as a bivariate normal distribution (two orthogonal independent bell curve distributions combined to provide X-Y impact positions) has been done since long before computers. A read of Hatcher's descriptions of group evaluation shows the army was using the assumption of a bivariate normal distribution in the 1920s. The Circular Area Probable (CEP) used in artillery targeting is based on the bivariate normal distribution. For more information on the principle and how it relates to group size, the ballistipedia article on it is good.
The bottom line is that group scatter was observed to obey two-axis normal distribution behavior long before computers were invented. Its use in computer models is merely adapting old knowledge to model expected typical group scatter. As long as you are interested in typical behavior, it works fine.
I was loading some h110 loads that were over book. They seemed safe, but were basically un shoot able to me. So, I’m loading about book min. Sure, many would argue with using h110 or loading a bullet without published data. Do you know of data for Bear Creek Supply bullets….me either. When you call Hodgdon, they say start with the weight and work up. Well, at 260gr, they show 24gr. I’m loading 24 gr at 265gr. It seems like I’m over…but I’m shooting lead bullets instead of copper jacket. Lots of published data at 260gr with cast bullets up to 27gr.
I agree. I would likely buy 2400 or AA#9 powder if it was growing on trees around here! H110 is decent, if it builds enough pressure to burn. I agree, it would make terrible powder in a 22000psi load!
The lead you see is from the bullets being pulled with a collet pulling tool. These are bullets I pulled deciding they were too hot for me.
You can see I have a lack of crimp? What is a good crimp? I set my seating die to just take the bell out. Then I crimp with a profile crimp die. It seems to keep the bullets from moving under recoil.
Wow! Are those 20 shot groups or something less? You sir are an excellent shot. I doubt I could pull that off with my 5.5” Blackhawk ever! Well, I’ll keep practicing! I’d like to get 5 into 6” at 50 yards. I can dream!