Othello的Minimax算法无法正常工作

Question

public class OthelloJPlayer extends OthelloPlayer {
  @Override
  public OthelloMove getMove(OthelloState state) {
    int bestchoice = 0;
    int bestscore = Integer.MIN_VALUE;
    boolean maximizingPlayer = true;

    // generate the list of moves:
    List<OthelloMove> moves = state.generateMoves();

    if (moves.isEmpty()) {
      // If there are no possible moves, just return "pass":
      return null;
    } else {
      // turn moves to states
      List<OthelloState> states = new ArrayList<OthelloState>();

      for (int i = 0; i < moves.size(); i++) {
        states.add(state.applyMoveCloning(moves.get(i)));
      }

      for (int i = 0; i < states.size(); i++) {
        // uses minmax to determine best move.
        int score = (MinMax(3, states.get(i), maximizingPlayer));

        if (score > bestscore) {
          bestscore = score;
          bestchoice = i;

        }
      }
    }

    return moves.get(bestchoice);
  }

  // min max algorithm
  public int MinMax(int depth, OthelloState game_board, boolean maximizingPlayer) {
    List<OthelloMove> moves;

    if (depth == 0) {
      int score = game_board.score();

      return score;
    }

    if (maximizingPlayer) {
      int bestvalue = Integer.MIN_VALUE;
      // gets other players moves
      moves = game_board.generateMoves(1);

      if (moves.isEmpty()) {
        int score = game_board.score();

        return score;

      } else {
        for (int i = 0; i < moves.size(); i++) {
          OthelloState new_game_board = new OthelloState(8);
          new_game_board = game_board.applyMoveCloning(moves.get(i));

          int returned_score = MinMax(depth - 1, new_game_board, false);
          bestvalue = max(bestvalue, returned_score);
        }
      }
      return bestvalue;
    } else {
      int bestvalue = Integer.MAX_VALUE;
      // gets your moves
      moves = game_board.generateMoves(0);

      if (moves.isEmpty()) {
        int score = game_board.score();

        return score;
      } else {
        for (int i = 0; i < moves.size(); i++) {
          OthelloState new_game_board = new OthelloState(8);
          new_game_board = game_board.applyMoveCloning(moves.get(i));

          int returned_score = MinMax(depth - 1, new_game_board, true);
          bestvalue = min(bestvalue, returned_score);
        }
      }

      return bestvalue;
    }
  }
}

My minimax algorithms does not appear to be returning the most optimal move. When my agent that uses minimax agent plays against an agent that performs random moves sometimes it looses. from my perceptive everything looks okay could someone please check my logic I must be missing something. The heuristic is the score. a positive score means you are winning a negative score means the other player is winning.

Answer 1

You have a number of issues.

1. Your getMove method is really the root of the search, and it's a max node. Therefore, it should invoke MinMax with maximizingPlayer = false .

2. When you invoke MinMax , you need to swap players. Right now, you just go from max -> max -> min -> min -> min... since you use true and false constants. Change your invocation (for both the min and max cases) to MinMax(depth - 1, new_game_board, !maximizingPlayer) .

3. Make sure that game_board.score() gives an evaluation from the perspective of the max player.