The best information on analysis of a shot group's found in this website:
Group Size Analysis
Simply overviewed, extreme spread's the worst and either mean radius or radial standard deviation's the best. This applies to finding the point in a ladder where a given load prints shot holes in elevation as well as what load produces the best accuracy.
For example, US military arsenals measure small arms ammo accuracy by mean radius. They shoot several dozen shots per test group. I've seen as many as 200 in a 600 yard group of 7.62 NATO ammo. Yup, 'tis hard to pinpoint the center of some of those holes in the middle of a 10-inch group of them, but the results are almost 100% reliable in assesing what the worst accuracy a given lot of ammo shoots.
Worst accuracy is always the best to claim for small arms ammunition. When one bases accuracy on the smallest group(s) shot, they're using a very tiny percentage of all groups/shots fired. Virtually all the others will shoot worse. Sometimes much worse.
Someone told me years ago to decide what accuracy level I wanted by defining what's the greatest distance I would accept missing my point of aim at the target's range. Then multiply that distance by two to get the worst accuracy that would meet my objectives. For example, if I wanted to miss a 500 yard point of aim by no more than 2 inches, I'd need a rifle and ammo that would shoot no worse than 4 inches at 500 yards; that's 8/10ths MOA. Statistacally speaking, that would require stuff that shot a mean radius at 500 yards of about 1.3 inches; that's a group average of about 2.6 inches.