Other Software > Developer's Corner

Lines Of Action Game Heuristics

(1/1)

jgpaiva:
My Artificial Inteligence work just came out, and I must say I'm abit scared. I have to make a computer player of the game Lines Of Action. I already know which algorithm to use, my problem is that I've never played the game, so, I don't quite know how to teach a computer to do so.
I've already searched quite a bit, and found good information about the game, but nothing on how to play it, strategically speaking.
So, I'm asking for some help, I'd like to know if someone knows some ways of playing the game (i.e. good heuristics, technically speaking), or at least, some good pointers for me to dig through.

(I couldn't fit this post anywhere, so, i posted it here because it's programming-related)

mouser:
game playing AI is one of my favorite subjects - my AI research is not in games but I still love it.

any idea what kind of approach you want to follow?

i.e. do you want to make a program that LEARNS to play:
http://www.research.ibm.com/massive/tdl.html

or do you want to try to design a playing algorithm yourself that doesnt actually learn.

jgpaiva:
I think I should have explained things a little better, but I'm also trying to figure out somethings myself.
I know the algorithm i'll use, it's minimax with Alpha-Beta cutoff. That's not a problem.
Another important issue is that my program can't store any data in between different games (although i'm not sure if it will between plays, I'm waiting an answer to that question).
I think it'll be quite basic, i don't think it'll take any logic deductions based on a knowledge base.
So, i guess there'll be no learning involved, because even if it could learn between plays in the came game, it wouldn't make much difference.
In the meanwhile, i've found a good link: The Quad Heuristic In Lines Of Action. It has some information about creating heuristics based on the center of mass and amount of "blocks", which makes some sense to me.

Navigation

[0] Message Index