Thread: Stupid Question. bare with me please

Originally Posted by Pommy

Needs a lot of rounds to give a good enough sample size to say with any certainty that a particular charge is giving a better ES than another. 3 rounds is not nearly enough. It probably won't make any meaningful difference. Same goes for grouping.

Agreed. What does a 3 shot group tell us about ES?


Well let's look at the ES of 15 different groups (ranging from 7-20 shots) vs the ES you get from the first 3 shots in the group. It looks like this.




The first 3 shots aren't very representative of the (more true) ES that appears with a larger sample in the majority of cases.


we can also look at a sample of 39 shots with the same rifle/load, sampled artificially as "3 shot groups" and see how representative of the true ES 3 shot groups are. It turns out that they are wildly variable; and none come close to demonstrating the "truer" ES measured across all shots.





However ES is quite a poor measure of how consistent our velocity is and how consistent we can expect it to be in future. It only takes into account the 2 most widely separated values and ignores the values of all other shots. Standard deviation is a better tool. Depending on the variability of your velocity, you should expect to get "reasonably" close to the true SD value across a sample of 10-15 rounds. Standard deviation is what should be used to assess hit probability.

4x your standard deviation (2 sigma) should be the range in which 95% of your velocities can be expected to fall. Note that I seem to have a sample size where that is currently true. I would expect 99.7% of my shots to fall within a range of 6x my SD - i.e. about 115-120fps "REAL ES". Note this is a .223 and every .223 load that I've collected sufficient data on, across several different rifles, brass, bullets and powders, loaded by different people, seems to end up with an SD around 15-20.

3 shot groups do not give you a good estimate of your SD. The standard practise of shooting 3-shot groups at different charge weights for load development and looking for a low value is invalid.

It can look like this:


You might conclude that load number 4 is the best load based on the consistent SD. There's a "node" there, right?

Wrong - every group on this chart is shot with the same load. The 3-shot SDs tell you essentially nothing.