Minimax 工作正常，但 Alpha-beta 修剪不能

Question

我試圖讓 Alpha-beta 修剪工作，但與我的 Minimax function 相比，它給了我完全錯誤的動作。 這是我的 Minimax function，它現在運行良好。

float Minimax(char[,] _board, int depth, bool isMax) {
    if (depth == 0 || isFull(_board)) {
        EvaluateBoard(_board);
    }

    if (isMax) {
        float bestScore = -Mathf.Infinity;
        float score = -Mathf.Infinity;

        for (int i = 0; i < 7; i++) {
            char[,] board = (char[, ]) _board.Clone();
            if (Play(board, i, false)) {
                score = Minimax(board, depth - 1, false);
                bestScore = Mathf.Max(score, bestScore);
            }
        }
        return bestScore;
    } else {
        float bestScore = Mathf.Infinity;
        float score = Mathf.Infinity;

        for (int i = 0; i < 7; i++) {
            char[,] board = (char[, ]) _board.Clone();
            if (Play(board, i, true)) {
                score = Minimax(board, depth - 1, true);
                bestScore = Mathf.Min(score, bestScore);
            }
        }
        return bestScore;
    }
}

這是我的 Alphabeta 修剪 function

float ABPruning(char[,] _board, int depth, float alpha, float beta, bool isMax) {
    if (depth == 0 || isFull(_board)) {
        return EvaluateBoard(_board);
    }
    if (isMax) {
        float bestScore = -Mathf.Infinity;
        
        for (int i = 0; i < 7; i++) {
            char[, ] board = (char[,]) _board.Clone();
            if (Play(board, i, false)) {
                bestScore = ABPruning(board, depth - 1, alpha, beta, false);
                alpha = Mathf.Max(alpha, bestScore);
                if (beta<=alpha) {
                    return bestScore;
                }

            }
        }
        return bestScore;
    } else {
        float bestScore = Mathf.Infinity;

        for (int i = 0; i < 7; i++) {
            char[, ] board = (char[,]) _board.Clone();
            if (Play(board, i, true)) {
                bestScore = ABPruning(board, depth - 1, alpha, beta, true);
                beta = Mathf.Min(beta, bestScore);
                if (beta<=alpha) {
                    break;
                }

            }
        }
        return bestScore;
    }
}

兩者都使用相同的評估，不確定這里有什么問題。 謝謝您的幫助。

Answer 1

目前尚不清楚您是在嘗試實現Fail-hard 還是 Fail-soft 方法，但我認為是后者。 如果是這樣，那么雖然我無法為您測試，但我認為這就是您想要的：

float ABPruning(char[,] _board, int depth, float alpha, float beta, bool isMax) {
    if (depth == 0 || isFull(_board)) {
        return EvaluateBoard(_board);
    }
    if (isMax) {
        float bestScore = -Mathf.Infinity;
        
        for (int i = 0; i < 7; i++) {
            char[, ] board = (char[,]) _board.Clone();
            if (Play(board, i, false)) {
                bestScore = Mathf.Max(bestScore, ABPruning(board, depth - 1, alpha, beta, false));
                alpha = Mathf.Max(alpha, bestScore);
                if (bestScore>=beta) {
                    break;
                }
            }
        }
        return bestScore;
    } else {
        float bestScore = Mathf.Infinity;

        for (int i = 0; i < 7; i++) {
            char[, ] board = (char[,]) _board.Clone();
            if (Play(board, i, true)) {
                bestScore = Mathf.Min(bestScore, ABPruning(board, depth - 1, alpha, beta, true));
                beta = Mathf.Min(beta, bestScore);
                if (bestScore<=alpha) {
                    break;
                }
            }
        }
        return bestScore;
    }
}