Hi,

I'm trying to create a simple AI for a board game that is similar to tic tac toe and othello, but in my game the board is almost infinate in size. I'm not yet sure how to tackle this, but it is ok that the AI bot is not as smart as compared to the AI for mentioned games. I've looked at Minimax and Alpha Beta Pruning, but no not sure if that is the right way to go in my case. Any ideas?

/Anders

Almost infinite? Is that like "a bit pregnant"?

Presumably the game has some constraints on where pieces can be placed, if they need to be connected you can run minimax etc. on it, just set the coordinates of the first piece played to be the origin and continue from there.

Otherwise a lot more details are going to be needed. First off you are going to need to tell us how you are going to represent the board.

Almost infinite? Is that like "a bit pregnant"?

I can imagine things I would describe as "almost infinite". For instance, the board is 2^64 x 2^64, but most of it is empty, so the board description is a list of the points of the board that are not empty.

Anyway, without knowing many more details about the game, it's impossible to tell what might work.

Think about how many options you have available per move (if the board is almost infiinite - then potentially you have almost an infinite amount of options), can those options be limited in some manner - are only a subset of them reasonable to inspect. If you can answer those questions it may lead you closer to an answer on how to solve it. For instance an almost infinite amount of viable options instantly rules out adversarial search like minimax/ab-pruning.

Just think about how you play the game, and write an AI to play the same way. Don't look at algorithms designed to use CPU resources efficiently, just try to make your AI human-seeming by using the same algorithms and strategies that you would tell your friends to use.

I don't think that's a good approach. Just because you play chess a certain way doesn't mean it's a good way to start building your AI. Start thinking about what you're proposing; could you even build a flow chart with a set of rules that could represent the way you think about any possible given situation? The answer to that question is no; not unless it's grossly simplified. If your AI decision making is grossly simplified, then it can't be reasonably expected to perform satisfyingly against a human player.

Chess AI is already quite the field of research, and generally speaking it's approached from an heavily mathematical perspective. I would strongly recommend understanding your game from the same kind of perspective before trying to build an appropriate engine to compute moves for it.