|
|
|
Discussion Forums
Forum Index
> Precision long range hunting and shooting
> Statistical significance
Statistical significance
@ 10:46 am (GMT) |
Scott StruifIn Honady’s latest technical video, their senior ballistician repeats the claim that a minimum of 30 shots at a target at a given distance is required to have a dataset that is statistically significant:https://youtu.be/UXm64Funxnw I’m curious what percentage of their sales is attributable to competition shooters, compared to hunters. Obviously, a competition shooter’s typical day practicing at the range could easily entail shooting 30 or more shots. But for a hunter, would such a regimen be meaningful? Assuming the 30-shot rule is true, a hunter’s range practice should consist of one cold-barrel shot a day for a month, from which group-size could be calculated. |
Replies
@ 10:17 am (GMT) |
Andrew MurrayRe: Statistical significanceWhile a tad off topic, the concept of statistical significance is wonderful. However it also somewhat academic and a bit of a wank.It is a scientific concept which is incredibly important when it comes to things that can be highly controlled like medicine trials or other biological interactions. We'd look at like the incidence of mesothelioma in those who worked/lived around asbestos mines to bring about correlations and draw links/conclusions. When it comes to shooting I'd say this is linked to bench-rest/F Class shooters who do tend to control more elements of their shooting beyond the capacity of a hunter. They will utilise incredibly heavy sleds as rests, I've seen utes set up as mobile reloading stations, employing wheel chocks and spirit levels to make the stations is as flat/stable as possible. The barrels and rifles they use beyond what could be considered useful in a field environment. But of course hunting and F-Class shooting is not the same and the comparisons are not helpful. So when it comes to statistical significance, the conditions in hunting are not replicable in a statistically significant way to be able to transfer such data. What can you control? Your loads, your rifle (accurising as much as capable), and your hold. The hold and ensuring you've got it nailed is the most statistically significant element that any shooter can engage in. Can't reload? Buy the same ammo each time (assuming it works in your rifle) Can't accurise the rifle? This may also lead to wider groups however if you cannot do this then you simply cannot do it. But you can work on your hold shooting position to make sure it's consistent. To be honest the concept of statistical significance sounds like marketing more than anything useful for a shooter. Why not make it 50? Or 100? Why not 1000? You can't really know how your rifle shoots until you've worn out 10 or so barrels right? But be sure to use X-brand powder/cases/projectiles/primers while you do it ;) |
|
@ 12:07 pm (GMT) |
Scott StruifRe: Statistical significanceHi Andrew. Thanks for your comment. It’s amusing to read a gun review where the author tries multiple 3-shot groups with various factory ammo, one of which shoots sub-MOA, from which he concludes the gun has met its sub-MOA guarantee.I wonder whether most hunters own a rifle capable of withstanding the rigors of shooting 30 or 40 rounds in a couple hours at a range. Most hunters don’t want to lug around rifles with heavy barrels. They opt for sporter contour barrels that competition shooters would scoff at. I agree that Hornady’s podcasts are aimed at the competition crowd. |
|
@ 01:01 pm (GMT) |
Andrew MurrayRe: Statistical significanceIt's always a good laugh when you go out and shoot sub 0.5MOA groups off a pack with a hunting rifle while blokes on the led sled and otherwise are doing the same or worse. 🤣 |
@ 07:18 pm (GMT) |
VinceRe: Statistical significanceHi ScottI agree that from a marketing perspective that target shooters are probably a more valuable resource 1:1 versus a hunter, although there are potentially more hunters out there, in terms of rounds down range, range shooting is probably where they make the most money. The cool thing about the statistically significant group quantities are that by and large they have done the leg work for us assuming the projectile you wish to use is in the 4doff database. The variables are of course shooter capability (up to the individual to solve regardless) and the specific rifle/scope /load combination but if we now have access to what is a corrected BC for a given projectile I'd call that a win. The problem with all of this for me is that for the ranges I currently shoot too and my own capabilities based on practice or lack therof, the difference in bullet impact is likely to be significantly less than my own margin of error.... Cheers Vince |
|
@ 07:17 am (GMT) |
Scott StruifRe: Statistical significanceHi Vince. I took him to mean that it takes 30 shots with the same load out of the same barrel to have a statistically significant dataset. But he also said they shoot each bullet out of multiple barrels, and then average the results to populate the 4dof dataset. That could mean shooting 5 shots each from 6 different barrels. Do they tune loads for each barrel, or use the same load? Either way, I’m convinced that their data saves the average hunter a lot of time and money over using a BC calculator. They said they’d do another video on the advanced 4dof features, so maybe I’ll post a comment asking them to explain further.After their 2nd video on drag, I was left with the impression that drag is primarily a function of the bullet’s shape. They also discussed gyroscopic stability, center of gravity, angle of attack, etc. I asked them why they put a polymer tip on their ELD-M match bullet, when they can draw or swage bullets into any shape they want. I pointed out that Berger does it that way, and that Berger’s 109 grain 6mm match bullet has less drop than a 109 grain 6mm ELD-M, according to their own 4dof data. They said, “There’s good reason for it,” but didn’t say what it is. Maybe it’s cheaper to manufacture a bulletin with a polymer tip. |
|
@ 09:29 am (GMT) |
Scott StruifRe: Statistical significanceHornady just posted another podcast on YouTube that answered my question. They said a 30 shot sample with the same barrel and load is optimal for a statistically significant sample:https://youtu.be/QwumAGRmz2I At the very end they said seating depth has a negligible effect on accuracy, compared to powder choice and bullet choice. They also said that seven 3-shot groups is the minimum dataset necessary to evaluate the capabilities of a hunting rifle. Further, that averaging the 7 group sizes doesn’t give you an average group size. You have to average the distance from the point of aim of all 21 shots. This all makes sense to me, except for the 3-shot groups for a hunting rifle. They said they chose 3 shots because most hunters can anticipate 2 follow-up shots. I would argue that's unrealistic. The only shot that matters for hunting rifle is the first cold-barrel shot. In a post a couple years ago, Nathan said he has a gong set up at long range that he can stop and take a shot at when he drives by. That makes sense. |
|
@ 09:41 am (GMT) |
Andrew MurrayRe: Statistical significance"At the very end they said seating depth has a negligible effect on accuracy, compared to powder choice and bullet choice."I recall them saying the opposite in that 7mm PRC video they posted. Not exactly that, but having lower powder charge for cartridge size can cause inaccuracies. Seating depth obviously changes the volume inside the case, thereby changing the way the burning powder interacts with the case. I could be wrong here (likely), but that sounds a strong contradiction within the space of 3-4 podcasts. My default setting is set to sceptic (ironic for a minister of religion) when it comes to manufacturers/marketers delivering information. A great example is Barnes doing their "shoot the milk jug at 2 metres with a solid copper vs lead projectile and compare the projectiles after" trick at the shooting range. But back to statistical significance: From manufacturing, I agree this is important. But for the shooter, less so. |
|
@ 10:43 am (GMT) |
Scott StruifRe: Statistical significanceThey said in the 7mm PRC video that they couldn’t neck down the 300 PRC because uphill and downhill shots would be inconsistent due to the reduced powder amount shifting back and forth.They said elsewhere that shooting reduced loads into gel at close range, to simulate long range performance, is useless because it fails to take into account gyroscopic stability. A bullet’s rpm, imparted by barrel twist, is directly proportional to its velocity, and doesn’t change much down range. So shooting a reduced load at close range means it could upset easier, due to tumbling caused by lack of gyroscopic stability. That’s why they don’t recommend a twist rate faster than the minimum necessary to stabilize the bullet. It’s worth keeping in mind though, that the point of their podcasts is marketing. |
|
@ 08:37 pm (GMT) |
VinceRe: Statistical significanceDoes this mean that when doing the statistical analysis of a 21 shot group minimum you should fire 7 shots 30 degrees uphill, 7 shots parallel and 7 shots 30 degrees downhill to make sure that your fill ratio isn't affecting POI, based on the potential in the field for angled shots?I kind of see the point they are making though, I'm tempted to give the bigger group size a try over Christmas with my triple deuce just to see what happens. Given that I shoot this suppressed, any ideas on how many shots at a time, should I clean at some point during shooting, all one target or 7 3 shot groups? Cheers Vince |
|
@ 08:07 am (GMT) |
Scott StruifRe: Statistical significanceThey said the gun has to be in the condition in which it’s going to be used prior to testing. So you’d need to know how many fouling shots it needs after cleaning. If you printed 7 of Nathan’s targets and shot 3-shot groups at each, it would be easy enough to lay each successive target on the first and mark the groups on it with a pencil.Obviously, if you shot a 0.5 MOA group on a target on a calm day, and a 0.5 MOA group in a different place on another target under similar conditions on a different day, you couldn’t declare your gun a 0.5 MOA rifle. That’s just common sense. I had no idea it would take 20 - 30 samples for a statistically significant dataset. And that’s just for the gun! Of course, the same principal could be applied to operator error. |
@ 03:12 pm (GMT) |
Florida_CrackerRe: Statistical significanceI don't want to rain on Hornady's party, but there is no significance (pun intended) over the statement that "30 observations" are required. It used to be taught in older texts as a "rule of thumb" but there are simple procedures using readily available data (standard deviation of a small sampling) that can be used to calculate the number of observations required to establish a chosen level of confidence. In general terms, the higher the SD, the more observations are required to be 95% certain (or other user defined level) that the results are not random observations. |
@ 03:33 pm (GMT) |
David LenziRe: Statistical significanceThere are a lot of folks that tend to engage in ballistic masturbation... they make noise, it's satisfying, and then they draw all manner of unsubstantiated conclusions in the afterglow (perhaps with cigarette lit).Much of my shooting background is tactically focused (job required) - plenty of folks train what might charitably be termed "theatrics" and they train past the point of diminishing returns. As a trainer of some note says: "you're either training to get better, or you're not." If what you're doing is yielding poor results, you're not training to get better. The limitations of the shooter and the equipment are always to be considered. 30 shots has been used by the US Army during lot testing for precision - and is generally considered to define the dispersion (precision) of a given system with high confidence. In other words, there is too much variability/uncertainty in smaller group/sample sizes to accurately gauge the absolute performance of a load (per contract spec) in a "one and done" type of test. Mathematically speaking, the average 30 shot group will not vary much, and it will be approximately twice as large as the average three shot group (which can vary a LOT, and at random, between groups). The recreational shooter and hunter is in a very different scenario. We can test continuously over the life of the barrel or as long as we own the gun if we are deliberate in our training/practice. Whether we track our cold bore shot every day (a technique I have seen executed more or less as Scott suggests and recommended by no less than White Feather himself), or aggregate the groups we fire, we enhance our understanding of both the rifle and ourselves. The trouble is our tendency to look for differences where we can't be sure any exist. Individual 3 (or even 5) shot groups are a TERRIBLE basis for decisions regarding load development and potential precision (unless differences are extreme). Ladder tests are often an example of this. I've done it, as have others. You can chase your tail all day (and guess at the results), or you can shoot a sample large enough to generate statistically significant data. The closer the performance, the more robust the data needed to differentiate two alternatives. Science, after all, comes down to "is it repeatable." As hunters or practically minded shooters, that's what matters, no? |
@ 05:19 am (GMT) |
Joshua MayfieldRe: Statistical significancePlease forgive a bit of a left turn... David brings up cold bore shots. Each rifle would be its own case, of course, and even one rifle would vary by load and environmental conditions, but is anyone aware of a good rule of thumb about how long one needs to let a rifle sit for the next shot to be, in true effect, a cold bore shot? |
Please wait
ABOUT USWe are a small, family run business, based out of Taranaki, New Zealand, who specialize in cartridge research and testing, and rifle accurizing.
|
KNOWLEDGEBASE |
KNOWLEDGEBASE |