带有 Minimax 的 Tic Tac Toe：当玩家先行时，计算机有时会丢失；否则工作

Question

我正在研究无与伦比的 Tic Tac Toe 的 Minimax 算法。我需要它在计算机先运行和播放器先运行时都能工作。使用当前版本，计算机在先行时永远不会丢失。然而，如果玩家先走，Minimax 似乎永远不会找到最佳移动（总是返回 -1 作为分数）。

如果玩家迈出第一步，是什么导致计算机的 Minimax 分数返回 -1？

例子：

board.addMark( 1, Mark2.PLAYER ); // add a 'PLAYER' mark to an arbitrary location
Minimax.minimax( board, Mark2.COMPUTER ); // will always return -1

这是“Minimax”类：

public class Minimax {
  public static int minimax( Board board, Mark2 currentMark ) {
    int score = (currentMark == Mark2.COMPUTER) ? -1 : 1;
    int[] availableSpaces = board.getAvailableSpaces();

    if ( board.hasWinningSolution() )
      score = (board.getWinningMark() == Mark2.COMPUTER) ? 1 : -1;
    else if ( availableSpaces.length != 0 ) {
      Mark2 nextMark = (currentMark == Mark2.COMPUTER) ? Mark2.PLAYER : Mark2.COMPUTER;

      for ( int availableIndex = 0; availableIndex < availableSpaces.length; availableIndex++ ) {
        board.addMark( availableSpaces[availableIndex], currentMark );
        int nextScore = minimax( board, nextMark );
        board.eraseMark( availableSpaces[availableIndex] );

        if ( currentMark == Mark2.COMPUTER && nextScore > score
            || currentMark == Mark2.PLAYER && nextScore < score )
          score = nextScore;
      }
    }

    return score;
  }
}

这是“董事会”类：

public class Board {
  private Mark2 gameBoard[];
  private int blankSpaces;

  private boolean solutionFound;
  private Mark2 winningMark;

  public final static int winSets[][] = {
      { 0, 1, 2 }, { 3, 4, 5 },
      { 6, 7, 8 }, { 0, 3, 6 },
      { 1, 4, 7 }, { 2, 5, 8 },
      { 0, 4, 8 }, { 2, 4, 6 }
  };

  public Board() {
    gameBoard = new Mark2[9];
    blankSpaces = 9;

    for ( int boardIndex = 0; boardIndex < gameBoard.length; boardIndex++ ) {
      gameBoard[boardIndex] = Mark2.BLANK;
    }
  }

  public void addMark( int spaceIndex, Mark2 mark ) {
    if ( gameBoard[spaceIndex] != mark ) {
      gameBoard[spaceIndex] = mark;

      if ( mark == Mark2.BLANK ) {
        blankSpaces++;
      } else {
        blankSpaces--;
      }
    }
  }

  public void eraseMark( int spaceIndex ) {
    if ( gameBoard[spaceIndex] != Mark2.BLANK ) {
      gameBoard[spaceIndex] = Mark2.BLANK;
      blankSpaces++;
    }
  }

  public int[] getAvailableSpaces() {
    int spaces[] = new int[blankSpaces];
    int spacesIndex = 0;

    for ( int boardIndex = 0; boardIndex < gameBoard.length; boardIndex++ )
      if ( gameBoard[boardIndex] == Mark2.BLANK )
        spaces[spacesIndex++] = boardIndex;

    return spaces;
  }

  public Mark2 getMarkAtIndex( int spaceIndex ) {
    return gameBoard[spaceIndex];
  }

  public boolean hasWinningSolution() {
    this.solutionFound = false;
    this.winningMark = Mark2.BLANK;

    for ( int setIndex = 0; setIndex < winSets.length && !solutionFound; setIndex++ )
      checkSpacesForWinningSolution( winSets[setIndex][0], winSets[setIndex][1], winSets[setIndex][2] );

    return solutionFound;
  }

  public Mark2 getWinningMark() {
    return this.winningMark;
  }

  private void checkSpacesForWinningSolution( int first, int second, int third ) {
    if ( gameBoard[first] == gameBoard[second] && gameBoard[third] == gameBoard[first]
        && gameBoard[first] != Mark2.BLANK ) {
      solutionFound = true;
      winningMark = gameBoard[first];
    }
  }

  public void printBoard() {
    System.out.printf( " %c | %c | %c\n", getMarkCharacter( gameBoard[0] ), getMarkCharacter( gameBoard[1] ),
        getMarkCharacter( gameBoard[2] ) );
    System.out.println( "------------" );
    System.out.printf( " %c | %c | %c\n", getMarkCharacter( gameBoard[3] ), getMarkCharacter( gameBoard[4] ),
        getMarkCharacter( gameBoard[5] ) );
    System.out.println( "------------" );
    System.out.printf( " %c | %c | %c\n", getMarkCharacter( gameBoard[6] ), getMarkCharacter( gameBoard[7] ),
        getMarkCharacter( gameBoard[8] ) );
  }

  public char getMarkCharacter( Mark2 mark ) {
    char result = (mark == Mark2.PLAYER) ? 'O' : ' ';
    result = (mark == Mark2.COMPUTER) ? 'X' : result;
    return result;
  }
}

如果有任何混淆，这里是“Mark2”类：

public enum Mark2 {
  BLANK,
  PLAYER,
  COMPUTER
}

Answer 1

经过长时间的休息后，在@GuyAdini 的回复的帮助下，我顿悟了。我编写了一个测试来计算 minimax() 返回的三个可能分数的出现次数。结果，它没有产生任何 0 的结果，这让我意识到算法需要将 0 视为可能的分数。

我最初将“分数”的初始化更改为最低/最高可能结果（-1/1）并与它们进行比较。但是，这禁止结果仅从返回的分数集中获取最低/最高值，而是还包括初始化值。这破坏了结果。

我添加了以下与“分数”的条件变化的比较：

|| availableIndex == 0

这迫使所有剩余的比较都针对属于返回分数集的值进行。

public class Minimax {
  public static int minimax( Board board, Mark2 currentMark ) {
    int score = 0;
    int[] availableSpaces = board.getAvailableSpaces();

    if ( board.hasWinningSolution() )
      score = (board.getWinningMark() == Mark2.COMPUTER) ? 1 : -1;
    else if ( availableSpaces.length != 0 ) {
      Mark2 nextMark = (currentMark == Mark2.COMPUTER) ? Mark2.PLAYER : Mark2.COMPUTER;

      for ( int availableIndex = 0; availableIndex < availableSpaces.length; availableIndex++ ) {
        board.addMark( availableSpaces[availableIndex], currentMark );
        int nextScore = minimax( board, nextMark );
        board.eraseMark( availableSpaces[availableIndex] );

        if ( currentMark == Mark2.COMPUTER && nextScore > score
            || currentMark == Mark2.PLAYER && nextScore < score
            || availableIndex == 0 )
          score = nextScore;
      }
    }

    return score;
  }
}