Seems to me this is the basic difference between a search and situational approach to evaluating game position. Backgammon is another great example, and one tha

Seems to me this is the basic difference between a search and situational approach to evaluating game position. Backgammon is another great example, and one that I think has a higher branching factor than even Go does. Minmax only works when you can construct a tree of possible outcomes that is reasonably searchable. Many "contests," like backgammon and Go, cannot be represented as a tree of outcomes so neatly. So you have no choice but to fall back on algorithms that evaluate the current positions without full knowledge of possible outcomes, i.e. heuristics, neural nets, etc.