AI for puzzle games,
Members - Reputation: 158
Posted 22 April 2000 - 05:11 AM
Basically work out a heuristic for each aspect of the game, in Chess this would could be 1 point for a pawn, 5 points for a castle 10 points for the queen etc. and calculate all possible moves from the current position (or world state) then calculate all the possible moves from each of those world states and all of the possible moves from those world states and so on. You could have a cut of for the depth of the search or you could only expand the best five world states from any node in the search to speed things up.
Then find the best move to make based on what could happen in the future. The best move could be decided to be the best average score from all the potential world states possible from making that move or it could be the move which has the highest, lowest possible score i.e. for each potential move find the lowest heuristic score that can occur from this move and make the move which can potentially cause the least damage presuming the human opponent makes the best moves on their turn.
That''s basically the best method where CPU time is not a problem (IMHO).