Well, there's some enginered ways to do it. Being more of a hobbyist doing an audience perspective capture, I tend towards the group. If the group is in a semi circle of acoustic performers (un-amped / no stacks). I tend towards > 1/4 in from the end, but < 1/3 in from the end. Generally speaking there's 4 voices in my groups and that puts the mics pointed at the break between sections, ignoring the one down the center. If I'm too far from the group then I just point them both parallel and at the group. As best as can be determined through thick windscreens anyway.
If you have a more commercial interest, you're probably better off sticking with the engineered ways (ORTF, X/Y, M/S, .......). As those might lend themselves better to being mixed to mono (AM Radio / TV) and other processing effects commonly done for broadcasts. Although I'm sure this subject has been discussed before in greater detail than I could offer.