This is a bit of tangent from Troy’s comment, but I see most gaming experiences as being a lot like a four-legged stool.
A stool with three legs will fall over if any of the three legs are removed — which isn’t the case with most games.
With RPGs, three of the legs on the stool have to do with the game itself. The post Troy commented on discusses three possible legs, believability, interest and balance, but there are lots of potential legs. (They’ll vary somewhat by group.)
The fourth leg, though, is consistent: Hanging out with your friends. In my experience, most games can weather the loss of one leg, as long as it’s not that one.
For example, if a game has too much intra-party conflict for my tastes (effectively knocking out one leg of my ideal gaming experience), that’s okay — I still have the fourth leg, spending time with my friends, which makes any game fun.
Do you agree with this (very loose) theoretical model? Does it have any value as a tool for thinking about GMing, or about gaming in general?