Minimax algorithm
| MINIMAX-DECISION calculates
a best move in a ply (two turns by opposing players). It assumes that it is moving for the first
player, and thus wants to maximize its utility value for its move. |
|
| In odd levels of the tree
the MAX-VALUE function is run, and at even levels, the MIN-VALUE function is
run. |
|
| Calculation in the tree runs
in a depth first manner until we reach a leaf (either at an even or an odd
level) and the values are then propagated back up successively higher levels
of the tree. |
|
| The either tree (initial
state to the leaf nodes) must be searched before MINIMAX returns a
decision. As this is not a realistic
possibility (except for trivially small games) we must change this initial
version of the algorithm such that decision can be made in reasonable (e.g.
real) time. |