Leandro Saccoletto
Member
Hi guys,
I'm a puzzle match 3 that swap pieces similar to Puzzle League. One problem that I'm have is how to approach to CPU AI.
The game has a grid 7x11. At first I thought in do in "brute way". Like do a simulation test one of 66 possible moves, and in each outcome test another 66 moves, and repeat until I have 3 moves in advance. An find the best move , based a value (blocks removed, number of combos and chains). But the number of possibilities can be at max 66x66x66 = 287497 grids created, taht could extensive to a CPU.
I tried study chess and Tetris AI, that has some interested concepts of MiniMax (that look closer of what I did), genetic functions. But I think that arcarde games from Neo Geo and others (that has lower CPU power), it should have a best approach, but still uknown to me.
The AI that I'm seeking is to give a variable challenge to player, that somehow I could plan some sequence of chains.
I'm a puzzle match 3 that swap pieces similar to Puzzle League. One problem that I'm have is how to approach to CPU AI.
The game has a grid 7x11. At first I thought in do in "brute way". Like do a simulation test one of 66 possible moves, and in each outcome test another 66 moves, and repeat until I have 3 moves in advance. An find the best move , based a value (blocks removed, number of combos and chains). But the number of possibilities can be at max 66x66x66 = 287497 grids created, taht could extensive to a CPU.
I tried study chess and Tetris AI, that has some interested concepts of MiniMax (that look closer of what I did), genetic functions. But I think that arcarde games from Neo Geo and others (that has lower CPU power), it should have a best approach, but still uknown to me.
The AI that I'm seeking is to give a variable challenge to player, that somehow I could plan some sequence of chains.