I have seen lots of targets posted on the internet showing impressive shot groups. Some show 3 shot groups, some show 5 shot groups and so forth. Often, while shooting my 6 shot revolver, I'll group 3 or 4 shots pretty good and then the next 2 or 3 will open it up. Should I quit while I am ahead and take a picture, or must I fire all 6 shots?
You put your finger on one of the major problems of using group size (i.e., maximum spread) as a measure of accuracy (precision, really, but that's another discussion). As is obvious, 1-shot groups will always give you the minimum group size, and as you add shots to the group it can never get smaller, only larger. For a single gun/load combination, even under controlled conditions, the larger the group size you use as your test of accuracy, the larger those groups will be. This is both conceptually and statistically unsatisfying.
A much better way to evaluate accuracy is to use the average deviation, which is simply the average distance your shots fall from the calculated "center" of the group. It does involve a little math, but only a little. One convenient way to do it is to use one of those targets that have 1-inch squares around an aiming point. After firing a number of shots - the actual number doesn't matter a great deal, but more is better and you should try to fire at least a half dozen or so - measure the distance of each shot from the center (aiming point) in terms of the x,y grid of squares. Thus each shot has an assigned coordinate in the x,y system. Shots to the left of center will have a negative "x" and shots below center will have a negative "y" - that's not a problem, but if you'd rather not work with negative coordinates you can choose an origin (0,0 point) below and to the left of all the shots, which will then all have positive coordinates. The results will be the same.
The group center in your x,y system is then calculated as the average of the x's and the average of the y's. The distance of each shot from that center is given by our old friend the Pythagorean Theorem as the square root of the (difference between the two x's squared plus the differences of the two y's squared). This is much easier than it may sound, especially if you can work in Excel. The average of all the distances in the group is the average deviation, which is your measure of accuracy.
This value is an inherent property of the gun/load combination, as well as the usual other factors, and is independent of the number of shots fired. As you increase the size of the group, the estimate improves and can become either smaller or larger. Also, one "flyer" (which should usually be included in the calculation, unless you know it's not representative, e.g., maybe you sneezed and squeezed that shot off accidentally) doesn't have as much of an effect on the average deviation as it would if you were measuring maximum spread. Assuming the center of your group is not exactly at the aiming point, the calculated actual center lets you know exactly how much you need to move the sights, whether you're interested in zeroing at that distance or want some known amount of holdover.
One slight disadvantage of this method is that it takes a bit of getting used to in terms of evaluating what's "good." We've all come to think of 1 MOA maximum spread as being a bit of benchmark for an accurate rifle, but that number holds no particular significance using this method.
This method is (somewhat) related to the military's Circular Error Probably (CEP) method of evaluating accuracy, a good discussion of which, including some useful tools, can be found
here.