English
资源说明：Unbeatable AI for Tic Tac Toe
tic-tac-toe AI (man1)
Credit where credit is due: Xeda112358 suggests this method of representing
board state at http://www.omnimaga.org/index.php?topic=13070.0; I've built
on his suggestions to more efficiently and reliably win or tie almost any game.
In this state representation, a large matrix represents each possible position
(the rows) and which possible win states it contributes to (the columns, 1 if it
contributes, 0 if not). Throughout a game, an 8 entry array is maintained; when
the user plays a space, the row in the matrix corresponding to this space is added
to the game array, and when the computer plays the corresponding row is subtracted.
Using this state representation, it's easy to identify when the computer wins (a
-3 in the game state array) or when the user wins (3), or when one is about to
win (-2, 2), etc.
Using this, the computer prioritizes certain actions. First, it will win if it
can. If it can't win, it will block any attempt by the user to win. If there's
nothing to block, it'll first try to create a split (a setup along two possible
win configurations), or at least a single setup. If none of these more advanced
strategies can be followed, it will prioritize the center, then the corners,
then the middle edges. It will also defend against both types of splits (attempts
to set up a win in two directions), those using the edges of the board and those
using the diagonal. I'm pretty sure this AI is unbeatable, but if you find a way
to conquer it, please let me know!
Throughout the program, cells on the board are referenced like so:
y = 2 | 6 7 8
1 | 3 4 5
0 | 0 1 2
|--------
x = 0 1 2
本源码包内暂不包含可直接显示的源代码文件，请下载源码包。
