5rd groups aren't statistically significant! Wanna bet?!!
- Thread starter JR1200W3
- Start date
I hear you, I was trying to understand Hornady's view with that last sentence. Not saying I understand it myself or agree.
My ugrad was in a brutal Bio program whose professors hated stats and called stats use in Bio a manipulation and not true science. Main reason, sample sizes too small for the complexity of what is allegedly tested. Too small for various reasons usually related to experiment design. That was 40 years ago and it stuck with me like glue ever since.
I bet you're good at scolding your students, lol. I'm just kidding. I appreciate you taking the time to break this down for me. I certainly have a lot to learn about statistics and was misusing estimator in too broadly a way.
We discussed doing an ANOVA yesterday because the T test was just a hasty look that didn't require breaking out graph pad, but once we got the data it didn't seem necessary because it doesn't appear that a more thorough test could change the conclusion.
I think I'm going to start collecting individual 5rd groups each time I go out and dumping them on here. IF I just collect a long pattern of reproducible means and single digit SDs, low teens ES then the large aggregated data of them all can confirm that the 1st individual 5rd, and 2nd 5rd,...and 3rd, and 4th, etc aren't just random chance.
If you have prior experience with the rifle and load and you are going to weigh that pervious experience against the group you shoot, then when you shoot a 5-shot group, the sample size is not 5. It's 5+previous. So suggesting that 5 is enough is a bit silly.
Even though you have a high-performing system, you 100% will be better served by 20-30 shot sets for average MV, MV SD, and MPOI for zeroing. I don't know if this thread is pointed against myself or what has been said on the Hornady Podcast but our point has always been that you're time/money ahead to just knock out a single 20-30 shot string for that data collection to set up a ballistic solver profile and for use with hit probability calculations. After that, the system is tight. Just use it on targets until something falls off.
In your OP you mention "fliers" in the data and point to a reason for them. First off, I would say your reason for them is a hypothesis that needs more testing at best. So the fliers are not "explained away" by a hypothesis. And I will tell you from experience that those fliers are present almost every time you do a large sample test and they are what make up the population. They are what make the sample more closely a match to the population. Omitting them is called, "Confirmation Bias" or "Cherry Picking".
As far as heat goes, I have done a single shot every minute and I have done 30 shots continuous as fast as I can. It depends on the cartridge and barrel, but generally speaking things in the .308 class and smaller are fine with heavy contour barrels to shoot 30 shots in a single string before heat starts skewing results. For sure 20 shots.
We can have difference of opinion on conclusion, but I would say your data set falls exactly in line with the data I have collected. Really nice data set, but you still see variation in the 5-shot strings for Average MV and MV SD, and the 30 shot SD is larger than the average of the 5 shotters. Pretty typical.
Even though you have a high-performing system, you 100% will be better served by 20-30 shot sets for average MV, MV SD, and MPOI for zeroing. I don't know if this thread is pointed against myself or what has been said on the Hornady Podcast but our point has always been that you're time/money ahead to just knock out a single 20-30 shot string for that data collection to set up a ballistic solver profile and for use with hit probability calculations. After that, the system is tight. Just use it on targets until something falls off.
In your OP you mention "fliers" in the data and point to a reason for them. First off, I would say your reason for them is a hypothesis that needs more testing at best. So the fliers are not "explained away" by a hypothesis. And I will tell you from experience that those fliers are present almost every time you do a large sample test and they are what make up the population. They are what make the sample more closely a match to the population. Omitting them is called, "Confirmation Bias" or "Cherry Picking".
As far as heat goes, I have done a single shot every minute and I have done 30 shots continuous as fast as I can. It depends on the cartridge and barrel, but generally speaking things in the .308 class and smaller are fine with heavy contour barrels to shoot 30 shots in a single string before heat starts skewing results. For sure 20 shots.
We can have difference of opinion on conclusion, but I would say your data set falls exactly in line with the data I have collected. Really nice data set, but you still see variation in the 5-shot strings for Average MV and MV SD, and the 30 shot SD is larger than the average of the 5 shotters. Pretty typical.
For almost all of us we can blame our K-12 schooling. I didn't really become the sour skeptic I am today until college and grad school both forced me into lots of critical thinking. Too much, in fact, according to my family. They are "average folks" compared to me as they didn't go through the Must Think Critically Always post-HS classes I did. So they don't like challenging what they already believe or think they know. Whereas I can't trust anything I encounter!
I started listening to one of their podcasts and they were bitching about their comment section saying if you don’t like their point of view, don’t listen. So I turned it off and never went back.
See, here you showed the spirit of being helpful and elevating the discussion. You passed a test today.
I will still deduct points for the last jab at the statisticians. The good ones are worth their weight in Rhodium and not to blame for the folks who just fake being a statistician while having corrupt ethics and ruin the world for all of us. Sorting them out is the challenge.
We are being helpful to warn the rookies that there is no joy in burning 30 rounds when they are beating a dead horse, just like we are being helpful when we show the origins and history of scientific methods and statistics used for ballistics and it typically takes 30 samples to close in on a normal distribution.
There is nothing wrong with using smaller sample sizes and taking some risks in the learning curve when we admit there is a risk that going forward might show a gun and load sucks.
The real goal for a reloading forum, is teaching the no-maths how to make better choices with limited resources.
Several of your points are golden. Don't waste resources on low quality junk, regardless of the price. Don't kid yourself with cheap shortcuts and expect to put in your practice and range time. Knowing that their rig is telling them to make a change, versus stop wasting time in unproductive load development loops and go practice in winds and weather.
It is good to highlight the problems with both the economics of the learning curve and load development risks, as well as teach the origins of the math, science, and statistics. The whole point of the forums is to help the rookies if you think about it.
Carry on.
What I'm referring to is the willful act of choosing a belief, not because of critical thought, but to align yourself and subscribe to an identity group. I think this has been fomented in our society through social media and polarizing issues like COVID and national politics. It's more about rejecting others beliefs because fuck them. You see yourself at war with this other side and you're going to dig your heels in. You also want to belong to a particular identity group and you're signaling that.
This summarizes why my Bio profs hated stats. Because too many experiments were run, and submitted for publishing, and even got published, with lousy use of stats. Not using them properly to a pure scientist = discardable and even possible corruption of the experiment analysis process (what to fix next time around).
I completely understand the purpose of Stats and used properly they're an excellent tool, as a hammer is for nails. But hammers suck at phillips head screws, eh? (Misuse of stats there, analogy).
This happens for the same reasons I said. Take a minute to ponder it, see if you don't agree.
If you don't get foundational critical thinking skills in your K-12, and if you go past HS and still don't get them, you go through life lacking the practice of discernment -- unless maybe it's part of your job and you have to do it there, but often that gets treated as a task and not constant process.
These points are why some of the published research in science cannot be replicated. Along with small samples and too large of type I error rates. Look at the 6 sigma type I error rate Physics community uses. They don’t have a replication issues. All of these points are why some meta analysis are garbage (since it was brought up earlier). How many meta analysis's are done in nutrition science and then turned over by another meta analysis to be turned over by another meta analysis. I don’t know the number but it’s happened.
I’m sure what I said above applies to your professors distrust for stats
The entire argument about being "statistically significant" depends entirely on perspective.
For a benchrest shooter they have their stuff down to a science. How various environmental conditions impact the load, the rifle, and bullet flight results in different loads being used in the morning vs afternoon. A lot of the time the entire competition is literally a 5 shot group.
At the end of the day those guys are very in tune with how X impacts Y or Z and they try to counter act those things.
Hornaday on the other hand has very minimal interest in such a thing. A company like that is worried about commercial ammo sales and are focused on people buying their ammo. If they are doing a run of 50,000 rounds what is statistically significant to them is entirely different.
To them, those benchrest guys are an outlier. They are not interested in tracking what is humanly possible on an absolute scale because the factors involved change from hour to hour. The benchrest guys show all the time what is possible but to Hornaday that's not really commercially viable.
They in turn look at the entire thing premised on (for example) hunters that don't reload and are using only commercially available ammo.
Basically the entire argument is comparing apples to oranges.
For a benchrest shooter they have their stuff down to a science. How various environmental conditions impact the load, the rifle, and bullet flight results in different loads being used in the morning vs afternoon. A lot of the time the entire competition is literally a 5 shot group.
At the end of the day those guys are very in tune with how X impacts Y or Z and they try to counter act those things.
Hornaday on the other hand has very minimal interest in such a thing. A company like that is worried about commercial ammo sales and are focused on people buying their ammo. If they are doing a run of 50,000 rounds what is statistically significant to them is entirely different.
To them, those benchrest guys are an outlier. They are not interested in tracking what is humanly possible on an absolute scale because the factors involved change from hour to hour. The benchrest guys show all the time what is possible but to Hornaday that's not really commercially viable.
They in turn look at the entire thing premised on (for example) hunters that don't reload and are using only commercially available ammo.
Basically the entire argument is comparing apples to oranges.
I used to have some faith in "the system". That was when peer reviewed work meant that there was a good chance to move the needle forward in the world.
Then, several years back, I noticed a bad trend in the world where "peer review" was corrupted.
"Paper Mills" and "review mills" began to appear with regularity. They ran away with speech about "follow the science" when it really meant do what we are telling you because we control the science. Big Pharma and politicians got greedy and ruined my faith in "the system".
Never in recent history have we seen so many hundreds of so called peer reviewed white papers and journal articles get exposed as fraud and retracted. If it continues, we end up in the dark ages again. Let us hope for better.
JB.IC, you remember University of Utah, Pons & Fleischmann and "cold fusion" and impossible replication of their results, eh?
In Bio circa mid 1980s there was a lot of confirmation bias in stats use, which is misapplication or misuse to those that know pure Stats as a part of pure Maths. Stats are supposed to be detached, uninterested, just analyzing data.
RegionRat, we posted same-time. Your post is like what I was trying to say. My Bio profs saw the "paper mills" approach back in the 1980s.
In Bio circa mid 1980s there was a lot of confirmation bias in stats use, which is misapplication or misuse to those that know pure Stats as a part of pure Maths. Stats are supposed to be detached, uninterested, just analyzing data.
RegionRat, we posted same-time. Your post is like what I was trying to say. My Bio profs saw the "paper mills" approach back in the 1980s.
Well there was some issues in cosmology as well, but I guess the point was the Physics community finds out issues pretty quickly. They police their researchers. Unlike say the behavior science/psychology (some of those researchers are getting caught committing fraud left and right), nutrition, and medicine/public health sciences.
For sure - physics being maths heavy it's the best place to sort out the bogus stats.
In Bio mid-80s the problem arose in an area that is quite relevant today. Cell and molecular bio, and mostly, research aimed at pharma products. My profs warned us students about the prevalence of confirmation bias arising from the lucrative grant money offers.
In Bio mid-80s the problem arose in an area that is quite relevant today. Cell and molecular bio, and mostly, research aimed at pharma products. My profs warned us students about the prevalence of confirmation bias arising from the lucrative grant money offers.
Well, unfortunately you failed the test because this post was as pompous and condescending as your last. So I'm not going to dignify it with a legitimate response. Instead, you only get this from me,
--------break--------;
You still have a Dandy trickler on your bench, don't you....
LOL, I did have and use one many years ago.... I still have it somewhere....
If I am being pompous or annoying you, it is not my intention.
I will slip back into the bleachers and cheer for the players on the field.... being retired has it's benefits and one of them is not needing to compete with or bother folks if it isn't being helpful. Cheers.
First everyone should read this article. That way everyone will understand what is meant by statistical significance.
articles.outlier.org
Consider two test, one has a mean of 2750 fps and a standard deviation of 5 (Group-1 ) and the other has a mean of 2760 and a standard deviation of 12 (Group-2). The question becomes "are these statistically different?". The standard test is Welch's T-test and we would normally want a 95% confidence that the difference is significant. If the test is for 5 shots then Welch's T-Test would have a p-value 0.1422 and the conclusion is the sample average of Group-1 and Group-2 is not big enough to be statistically significant.
On the other hand if the Groups are 10 shots, then the p-value is 0.03154 and the sample average of Group-1 and Group-2 is big enough to be statistically significant.
Why is the conclusion different? It has to do with probability in sample selection. Too few samples can result in results with unusually low or unusually high averages with extremely small or large standard deviations.
In terms of confidence interval, the 5 shot 2750 confidence interval is 2743.8 to 2756.2. That means that if the test was completed many times 95% of the means would fall between those values. If the group size is ten then confidence interval is 2746.4 to 2753.6 meaning that the higher shot count has improved our estimate of the true population mean.
Similar test and estimates can be made for standard deviation but it is not normally distributed. In this case small sample sizes drastically bias the results to the low side. This is because 68% of a normally distributed population lies within 1 standard deviation. The more samples taken the greater the chances of getting a sample outside the one standard deviation. Comparisons between test standard deviations are usually done with the F-Test although there are others as well.
What Is Statistical Significance & Why Learn It | Outlier
Learn what statistical significance means, why it is important, and how it’s calculated, and what the levels of significance mean.
articles.outlier.org
Consider two test, one has a mean of 2750 fps and a standard deviation of 5 (Group-1 ) and the other has a mean of 2760 and a standard deviation of 12 (Group-2). The question becomes "are these statistically different?". The standard test is Welch's T-test and we would normally want a 95% confidence that the difference is significant. If the test is for 5 shots then Welch's T-Test would have a p-value 0.1422 and the conclusion is the sample average of Group-1 and Group-2 is not big enough to be statistically significant.
On the other hand if the Groups are 10 shots, then the p-value is 0.03154 and the sample average of Group-1 and Group-2 is big enough to be statistically significant.
Why is the conclusion different? It has to do with probability in sample selection. Too few samples can result in results with unusually low or unusually high averages with extremely small or large standard deviations.
In terms of confidence interval, the 5 shot 2750 confidence interval is 2743.8 to 2756.2. That means that if the test was completed many times 95% of the means would fall between those values. If the group size is ten then confidence interval is 2746.4 to 2753.6 meaning that the higher shot count has improved our estimate of the true population mean.
Similar test and estimates can be made for standard deviation but it is not normally distributed. In this case small sample sizes drastically bias the results to the low side. This is because 68% of a normally distributed population lies within 1 standard deviation. The more samples taken the greater the chances of getting a sample outside the one standard deviation. Comparisons between test standard deviations are usually done with the F-Test although there are others as well.
Really? That's not believable. Is that why you're taking the position of someone to pass judgement in a pass and fail way? Awarding and deducting of points? The passive aggressive "Carry on" as if someone needs your permission to go about their day because now you're done talking.
Have you ever heard of the importance of how you see yourself vs how others see you? I bet you genuinely tell yourself that you're an elevated intellectual that stays above the fray of the mosh-pit elbows, sticking to the content, but then self-delude thinking that if you throw the elbows but in a quasi -intellectual way it's different. And the person you're talking to isn't smart enough to see through it so you have a defensible cover-for-action.
At this point you aren't contributing in meaningful way, you're just trying to ease your chapped ass for cathartic reasons.
Hell yeah. That's good shit!First everyone should read this article. That way everyone will understand what is meant by statistical significance.
What Is Statistical Significance & Why Learn It | Outlier
Learn what statistical significance means, why it is important, and how it’s calculated, and what the levels of significance mean.
Consider two test, one has a mean of 2750 fps and a standard deviation of 5 (Group-1 ) and the other has a mean of 2760 and a standard deviation of 12 (Group-2). The question becomes "are these statistically different?". The standard test is Welch's T-test and we would normally want a 95% confidence that the difference is significant. If the test is for 5 shots then Welch's T-Test would have a p-value 0.1422 and the conclusion is the sample average of Group-1 and Group-2 is not big enough to be statistically significant.
On the other hand if the Groups are 10 shots, then the p-value is 0.03154 and the sample average of Group-1 and Group-2 is big enough to be statistically significant.
Why is the conclusion different? It has to do with probability in sample selection. Too few samples can result in results with unusually low or unusually high averages with extremely small or large standard deviations.
In terms of confidence interval, the 5 shot 2750 confidence interval is 2743.8 to 2756.2. That means that if the test was completed many times 95% of the means would fall between those values. If the group size is ten then confidence interval is 2746.4 to 2753.6 meaning that the higher shot count has improved our estimate of the true population mean.
Similar test and estimates can be made for standard deviation but it is not normally distributed. In this case small sample sizes drastically bias the results to the low side. This is because 68% of a normally distributed population lies within 1 standard deviation. The more samples taken the greater the chances of getting a sample outside the one standard deviation. Comparisons between test standard deviations are usually done with the F-Test although there are others as well.
Did you apply an error rate correction to those CI's you state?
And it was this exact moment I knew we were all going to argue
IYK….you prob don’t K, and just think you K…..
Also I didn’t see your null hypothesis clearly stated.
And this encapsulates my understanding of the whole thing. I am not a statistician, I only play one on TV. Just kidding. I am a caveman electrician, Years ago, your scientists thawed me out of a glacier...
I have only the vaguest understanding of statistics, in spite of being good at math and even being good at math in certain subjects. I have memorized the for voltage drop per distance, thanks to a colorful mnemonic once taught to me.
Anyway, the relevance of a group depends on the job. I have watched the Hornady podcasts around this subject 4 times each.
Jayden Quinlan said that one should group a rifle for the job. For example, for a hunting rifle, 7 groups of 3 rounds each will provide the most benefit. Anything after that would have diminishing returns, really.
Why go through 100 rounds? I don't think anyone is going to need that. But it would definitely give a high percentage reliability of what can be expected. Which is still not what the rifle is going to do. Just a prediction.
But I am also reminded of a statement by Wade Chandler at TPH Podcast (Texas Predator Hunting.) He once shot a single cold bore shot, once a day for five days in a row. There was still dispersion and it was indistinguishable from doing a 5 round group at one sitting. I wonder if you did that 6 weeks and get a 30 round group of cold bore, would be any different or larger than the first week of results. In that case, the initial 5 round single shot might only have more significance if you repeated it, which belies the knowledge it was not initially valuable unless verified.
And only for that rifle and whatever load data.
How valuable is the dispersion? Depends on the job. With game animals, you have a target zone of a certain size that is effective. Opposed to competitions where the winner has very small dispersion. I am reminded of what Erik Cortina said in relation to this. He will test a rifle and if it is not doing 3/8" in 3 shots, then he is not interested because 77 more shots will not help.
I am also reminded of what I see with the Texas Plinking challenge of 1 MOA at 1k yards, and the follow-up, where the winner gets a chance to pattern a 10 round group and see if the can shoot a 10" group at 1k yards.
So, a few guys have been able to hit a 1 MOA and 1/2 MOA target. And still, at the second contest, cannot get the pattern in 10 inches. Which is to say that hitting the 1 MOA target at 1k yards does not mean that it is a 1 MOA rifle. Blame it on the wind. These guys have good equipment and I have seen where most of them have usable elevation and the wind just kicks their butts.
Which might sound like I am questioning what relevance the stats have, opposed to just hitting the target.
"The target doesn't lie." - Erik Cortina
Speaking of which, Hornady did another podcast on whether tuner brakes, like Cortina's EC Tuner help to reduce dispersion. What they found was that any device such as brake, suppressor, or the EC Tuner did produce smaller dispersion but they could not tell any difference between all three and adjusting the tuner settings did not produce enough difference.
FFS I think that whole P vs K argument went down a while back. I think I understood it then. It might have involved this controversial dude named Bryan Zolnikov. I think he goes by @Tokay444 but I could be wrong about that detail?And it was this exact moment I knew we were all going to argue
IYK….you prob don’t K, and just think you K…..
Also I didn’t see your null hypothesis clearly stated.
I hate to report:FFS I think that whole P vs K argument went down a while back. I think I understood it then. It might have involved this controversial dude named Bryan Zolnikov. I think he goes by @Tokay444 but I could be wrong about that detail?
THAT fight happens once a week in many people’s work life. My favorites are when two insufferable graduate students are having it. I definitely stir that pot! While fully understanding the phrase “they deserve each other”
Now back to the show!
I’ll beat Tokay444 to it…”Dunning Kruger”I hate to report:
THAT fight happens once a week in many people’s work life. My favorites are when two insufferable graduate students are having it. I definitely stir that pot! While fully understanding the phrase “they deserve each other”
Now back to the show!
So I calculated the 95% CI for the 30rd string. Stated this in the OP. Then pointed out that String 5 has a mean outside the CI. Later in another post, I use this as evidence of a statistical significance.
Post in thread '5rd groups aren't statistically significant! Wanna bet?!!' https://www.snipershide.com/shootin...y-significant-wanna-bet.7265325/post-12271717
JB.IC mentioned that I failed String 5 as statistically significant because I failed to adjust my P values. I then countered by pointing out that String 5's mean failed to fall within the 95% CI of the overall 30rd string and in post #23 he reaffirms this isn't accurate either because the 95% CI should be adjusted for Type I error rate corrections.
So in your example, you just calculated a standard 95% CI, yeah?
Don’t forget I then supported your conclusion after making the correct type I error adjustments with the appropriate model. 15 statistical test without any error rate adjustments is an incorrect practice.
Pls demonstrate how? Compare the mean, SD, and ES of the 6 individual strings against the 30rd string and explain what steps a shooter would take to do something meaningfully different. Could I change the MV in my ballistic calculator from 2686 to 2692? Sure. But would it be meaningful? Would that low MV change result in a miss in a field of targets ranging from 1 moa to 3moa, inside 1000yds? Perhaps the SD or ES differences then? How do we use SDs as shooters? We don't calculate a drop data using them? They just provide us reassurance that our loading practice and condition of our barrel are solid. If the results were extremely inconsistent and peppered with 20fps SDs one string and the 3fps SDs the next string you would doubt your Chrono, powder scale, neck tension, or bore condition, right? But look at my data. Every single SD is single digit. The ES's don't even represent a meaningful value that I can dial on my scope at 1000. And they're certainly sufficient in terms of reloading.
So what should I do with the observed 30rd string data that I'm not already doing with any of those 5rd string data?
Pls demonstrate where I removed fliers or outliers from my data set? I specifically noted that I did not see any outliers in my data and that all values are the result of random chance in a distribution, actually. I think you are responding to the part where I attempt to distinguish between outliers and random variation. I translate what those two things mean to the reloader and shooter. I'm not going to respond to natural variation in a distribution that is meaningless on the reloading bench or in the ballistic solver. But if I see a clear outlier, I'm going to identify the cause and attempt to rectify it. Fix it. So I can get to the type of tight data that I am reliably reproducing now.
You talk about explaining away, cherry picking, and confirmation bias, but you only speak to one take on outliers. You fail to mention that data sets can have erroneous values that researchers cull or correct for. Just to be clear, I did not cull or correct any of my values, but it certainly is a thing in experiments when input errors, flawed test equipment, or anomalies in the test subject occur.(Like the patient died before the second comparative portion is performed) A relevant example is what if I tried to repeat this data set with a shot out barrel expecting it to predict what a new barrel would do? That would be flawed data set, wouldn't it? So, to lecture someone on cherry picking while leaving out real cases for eliminating anomalies in studies is a confirmation bias.
I'll keep looking at this. I did consider that I collected the loaded rounds as they came out of the 550 and loaded them in the ammo box in that sequence and then shot them in that sequence so it could also be a result of scale drift, or bore condition changing from super clean barrel with minimal fouling to collecting additional copper.
I certainly don't have a photographic memory or the vantage point to know every data set you've seen but from what I remembering you posting here in the hide you point out > 1 moa 30rd groups and SDs in the teens. Yes, the overall SD in my data set did increase, technically, but not outside the range of the normal expected variance. From 6 to 7.9, lol
Maybe I just didn't connect the dots. So are you saying the point I made about String 5, Mean MV and the 30rd CI was valid?
Let me state it this way, for multiple comparisons for the MVs, string 5 had a significant difference from string 2 and 3. All else were null. I’d have to go back and read everything again for the 30rd CI.
I expect your collection of data will show very similarly to what we have already seen.
I once collected 44 separate 10 shot groups over about 8 months. 22LR with CCISV. Average 1.77”@100. Every single group I shot at 100 during that time.
pls say where I said you said you removed fliers from the data set... It's all really easy to dismiss after you already know how it performs.
Have a nice thread.
It very much seems like you implied it, by lecturing about fliers, explaining away, and cherry picking. If that's not what you were doing then what was the point of that tangent? Just random thoughts inapplicable to the results of the test? If you were not accusing me of cherry picking and I just misunderstood then I'll apologize and just say, "cool...cool, cool".
I'll take the fact that you can't answer my first question as admission that there is no difference. You fail to reject the null hyposthesis.
I’d like to invite a shooter that’s going to show me how all this shit is worth the trouble of mental masturbation. My reloading process vs theirs. I think things should show at 1000yds. I got that range. I can travel to somebody’s range too. I’d prefer they have a private range like me though.
I find that confusing. Hornady used to be heavier into hunting than competition but that has started to change a bit. They sponsor PRS matches, they came out with the Atip, Ledzep talks about this very subject in the context of testing his competition rifles and loads for action/ precision rifle competitions. You're making the same point Floxgal made that Hornady is taking this stance in the context of their manufacturing of bullets, but Ledzep is translating that here in the hide to individual shooters. The podcast wasn't directed at plant managers or production engineers. It was directed at shooters and reloaders. But hell, the dude is here . Ask him.
And my little hobo test here isn't directed at Hornady or Ledzep per say, it's all the other folks on here that hear Hornady 's message and choose to interpret it in their own way. Hence why I'm offering evidence of one set of conditions that suggests otherwise. And I would also say this set of conditions I'm referring to isn't a niche anomaly. I would make a guess it's pretty common to guys shooting high quality barrels, using proven components like Berger bullets, Hodgdon extreme series of powders, match primers, Alpha and Lapua brass, making good choices of reloading tools like an accurate powder scale and trickling system, controlling neck tension, and case neck lubricity. Pretty common set of instruments in this community. I'm not using $600 SAC dies or $1200 reloading press. It's really not that far out of the norm.
Here's a question for the crowd. Are my six 5rd strings unrepresentative and noticably different than what you get? Are you regularly jumping between 5 and 15fps SDs? ES's in the 30's and 60's? My data isn't that crazy is it?
It's not that confusing. They do make and sell reloading components for sure. However a bigger part of their company is selling ammo.
The point being is that to truly tune your ammo for not only each gun, but for each session with each gun is something a commercial volume based ammo company can't reasonably do and nor should they.
If you think of every single load variation a bench rest shooter might have 10 different SKU for just one single rifle. There is no way any company could manage that type of thing for commercial ammo offerings.
The question is does that type of absolutely insane ammo tweaking actually produce results? Absolutely it does. And for people in the know, it's pretty reliable and repeatable.
On the other hand any big company could not be that granular with their ammo offerings. It's just flat out off the table, hence why I originally said they consider it as an outlier in regards to any "study" conducted by them or anyone else that produces large amounts of ammo.
Most people do not reload. They just shoot ammo that they buy, but naturally they want what works best for them. They are not reducing powder charges or changing seating depth. They have a standard ammo and go for it.
The latter scenario is what I think Hornady (or any other ammo company) is studying. With that being the constant what is and isn't significant is determined using those parameters.
Like I said, those bench rest guys are not using static ammo whereas those other companies are.
Really and truly they are actually playing different games. One is trying to shoot laser beams in one hole across various days and conditions with zero consideration for terminal effects on a deer. The other is trying to create a product for sale that will be used by a hundred thousand different people in a hundred thousand different guns.
Last edited:
All that being said, my opinion is that regular shooters don't need to shoot huge groups to get meaningful data.
Determine what the purpose is and test everything accordingly.
That said I can't think of any individual shooter that would need to copy the procedures of a bigger company to get significant data relevant to them.
Determine what the purpose is and test everything accordingly.
That said I can't think of any individual shooter that would need to copy the procedures of a bigger company to get significant data relevant to them.
I fundamentally don't understand what your argument is against what I/we have said.
Here's what we've said:
- 5 shot groups do not accurately distinguish two or more loads from one another.
- 5 shot groups (even in good/best case systems, like yours) are ~50/50 for being up to or more than 0.05 mil off for MPOI vs. a larger sample size. (i.e. with most scopes you are going to be on the wrong 0.1 click a significant percentage of the time-- 35-60% of the time for most precision rifles)
- 5 shot groups do not repeat in dispersion
- 5 shot groups do not repeat in average MV
- 5 shot groups do not repeat in MV SD
All of those metrics (Dispersion, avg MV, MV SD, MPOI) are repeatable to a level that is well within the adjustment/resolution capability of the optic/shooter with 30-50 shot groups. There is no second guessing it at that point. 20 is a great compromise point where the vast majority of the time you'll be repeatable with occasional times you will not.
Our entire reason for saying to shoot more rounds is completely 100% under the assumption that you have not shot 200 rounds through the rifle with that load already with that information in your back pocket. That is my main problem with your argument on this post. If you have a history with the rifle, the sample size is not 5 rounds, it's much more than that.
If you have a better shooting system, the total variation will be less. No argument there. YOU DON'T KNOW THAT BEFORE YOU SHOOT IT THOUGH.
The entire point of shooting more rounds is to have an absolutely rock solid set of data to feed into an app to use as a predictive tool to calculate down-range trajectories and assess hit probability with real, accurate data.
Whatever you do with 5 shots, the data will be better and more repeatable with 20-30.
<<<Another perspective on this>>>
Purposefully screw up the system some how. Load it with a different powder that shoots 10-12fps SD on a 30 shot string. Load up 50 of each load (50 of the current load and 50 of this new load). Give the boxes to someone else and have them put 5x each into two magazines so you don't know which is which.
Shoot the two 5 shot groups and guess which one is which... Repeat this 10 times.
Will you always be able to tell which is which?
Now try it again with 30x of each into a single 30x group each.
Will you be able to distinguish them now?
Yikes - I’ve never seen someone gain so much confidence from a few weeks of a low level college course and a single course of fire at a range.
I’m with @Ledzep clearly shooting 20-30 rounds is significantly better to understand this data than 5. Your es i 150% bigger in the 30 round group than your first 4 groups.
I’m with @Ledzep clearly shooting 20-30 rounds is significantly better to understand this data than 5. Your es i 150% bigger in the 30 round group than your first 4 groups.
1) No. With my equipment and shooting/reloading consistency (I consistently stink, but I'm consistent at it), I'm seeing extremely similar results.
2) No. * On .308 case or smaller cartridges that I baseline single-digit SDs on. It is harder with big magnums. Obviously environmental factors (summer/winter come into play).
3) No. I've posted videos where I've shot back-to-back 10 shot groups @ 300 that were within 1/8" of each other, and multiple times have shot exactly repeat group sizes @ 650.
*****
Dudes will hate on my opinion. I've been around long enough to see trends come and go. Just a few years back, you were labeled as mentally handicapped if you didn't shoot a 10-shot ladder and 'look for flat spots'. Prior to that it was OCW. Over time, I'll bet we see more guys get tired of spending an afternoon shooting one large group and gravitate back to just trusting an aggregate of a few smaller ones.
As always, YMMV.
Again, I think I/we have been clear on this... As long as you maintain some sort of POA/POI reference, it's the same thing. 6x 5 shot groups is 1x 30 shot group if POI/POA relationship is maintained.
However, averaging the SD on 6x 5-shot groups is NOT the same as the SD of a 30-shot sample. The avg of the 5 shots will always be lower than the 30 shot combined.
1. This isn't all about you. Keep in mind it's also about people parroting you in their own interpretations. Who lack the nuance to understand there's a special condition to necessitate this large data sample on an unknown gun that was unspoken.
But ...even now, in this thread, I honestly feel like I hear the goal posts being moved in what you're saying. Here's why.
There's never been any caveat given about DOPE. Data on prior engagements. That's never been in the conversation until now. And like the other guy, ....in what world are we gathering meaningful data on the first 30 rounds in a gun? Of course we have previous data and history with the gun. It's our gun! Lol. And we get a certain number of rounds down the barrel before we start taking the results real seriously. So our understanding of the rifle, the load, and it's capabilities is always a constant in these discussions. So when you say broadly, without caveat, "A 30rd sample is the minimum amount to truly understand your gun", everyone assumes that's along with and on top of what they're already doing with their rifles. (Yeah, you didn't say that in your last post, but you've said it before. And so has everyone that parrots you). Here's another reason why. Despite what I just typed, you are also saying this by 30rd by giving the basic statistical truth that 30rds will give a better statistical truth than 5. I know, I agree, in a vacuum. But just that one sentence alone gets exported to people believing they need to shoot 30rd groups every time they go to the range. The basic, undeniable truth that a large sample size gives greater confidence....skips right over the consideration of what a shooter can practically use from that data set for a lot of people when they hear it.
Look at it this way. What if on your podcast you had said, "You really need at least 30rds to understand what your gun and load are capable of, but it doesn't need to be shot in one sitting or even on the same day". Then it would be a pretty unnoteworthy statement and everyone would have shrugged their shoulders and walked away. And it wouldn't have justified the click bait title of the video, "Your groups are too small". And honestly, I don't think that's even what you meant. I very much think you straight up meant to tell people that they need to lay down and shoot a 30rd group every time they want to see what their MV, SD, and group size looks like because anything smaller is a lie.
And to just play the same game, I agree! A larger sample size is going to provide better data. But I still maintain that "better data" isn't always usable and the cons of shooting 30rd groups everytime you go to the range is F'ing retarded.
Not technically correct that the average of the SDs will always be lower than a combined SD.
I ran a simulation study where the sample SD from 6 groups of sample size 5 were calculate, then averaged and the grand SD of all 30. There’s about 21.7% likelihood that the average of the 6 SDs would be greater than the grand SD with a 99% CI [21.3%, 22.0%].
I ran it with sample sizes as 50 and the likelihood jumped to 25.6% with a 99% CI [25.2%, 25.9%]. The main take away is that the SD distribution is highly variable compared to other distributions like the sample mean.
But, I think your general point is fair since what you claimed happens on average 74.47-78.3% of the time.
Last edited:
I did 'the maths in skool' boss, I understand sample size populations, and will never argue against the statistical importance of a larger population.
My point *outside the part you quoted* was to the relevance of the sample size of an aggregate vs large sample - as it pertains to the vast, vast majority of shooters.
Let's use your SD point... Is a ~4 FPS SD increase (30 shot result of OPs samples vs his lowest 5 shot sample) relevant to elevation adjustment at say 800 yards? We know the answer, it is about a half inch, or a little less than one-fifth of .1 mil.
Either way man, we're going around in circles partially agreeing, and partially not. There's enough animosity already here that I don't need to add to it.
Similar threads
