Any people remember enough statistics to figure out the power calculation to determine what sample size I need to find a significant difference in ES or SD. Maybe a difference of 5 SD would seem worth finding?
A difference between means of 5 SD is actually very large, so you don't need many samples in each group to run your test if you only want to detect that large of a difference. Assuming that you're planning to use the traditional t-test with an alpha of .05, and a power of .8 (both of which are pretty standard), you only need 2 samples in each group if the true difference between means is 5 SD. That assumes normality and homoscedasticity, which are probably reasonable assumptions for the type of data we're talking about, and also the same number of rounds in each group.
Determining sample size by establishing a minimum difference in terms of SD is probably not the best way to go about this because I'm not sure how you decide what's an "important" difference in SD to test for. A better way might be to think in terms of the difference between the mean velocities of the groups. Is a true difference of 5 fps important? I'd say probably not. 50 fps? Yeah, that probably is. Let's say, for the sake of example, that we decide that a real difference between the two mean velocities of 15 fps is what we'd like to detect. For the power analysis, we also need to specify what the SD is, which can be approximated with a little testing ahead of time. As an example, I'm looking at some data I collected for factory 250 gr .45 Colt loads - approx. 700 fps (the actual velocity doesn't enter into the calculation, only the difference in velocity between the two groups) with an SD of 7.76, so let's use that.
So, if I want to detect a "true" difference of 15 fps, with an SD of 7.76, and the other parameter values discussed above, the required sample size would be 5 (in each group).
If you want to try some of these calculations, I use a program called PS, which is available on-line at no cost from Vanderbilt University. Here's the link:
http://biostat.mc.vanderbilt.edu/wiki/Main/PowerSampleSize