MetalGod said:
Thanks UN , lets take that to the next step or keep it to group size . In load development and you’re shooting five shot groups in half grain increments . Can that be used in load development? I’ve heard many times you need to shoot 50 shot groups to know how accurate a load will be . What if one load shoots 2 moa at 300 yards and the other group shoots sub moa at 300 yards ( five shot groups ) . Can you conclude one is better then the other . Or is the shooter himself too big of a variable in the equation ?
Everything hinges on how much variation you get. If it is large, it takes more samples to be sure you have a difference from one to another load. If the variation is small, sometimes a few shots can tell you there is a difference worth having.
One trick I have used in the past is to use running averages. So, suppose I use three shots for each group in a ladder. I then use a running average of three groups. That is, I average groups 1, 2, and 3. Then I average groups 2, 3, and 4. Then I average groups 3, 4, and 5, and so on. This way, I am considering 9 shots at a time, which is much better than three. And if I see a pattern in that, I have much more confidence it is not just random chance at work. In this way, smaller group sizes can be made to present valid information if the load difference between them is not too large.
As to how many shots you need in a group to know what is actually happening, that depends on defining "actually happening". If you need to know ±10% it takes fewer shots than if you need to know ±1%. Basically, the higher the resolution you demand, the more shots you will have to take. How exact your determination needs to be depends upon how exact you need it to be. The more exact, the more shots you will need to fire per group.
Member Statshooter teaches statistics. He always shoots 30-shot samples. If you take samples of 30 you will find that a histogram only starts to show the shape of the Gaussian bell curve with about 30 samples. So if true certainty is your objective, then 30 will be your minimum. But if you only want to know to 90% certainty that one load is more accurate than another, it can often be done with much less. Just don't expect that to tell you exactly how much better.