Issue with MiniMax to Alpha

OK, my issue should sound pretty familiar to anyone who has played with board game programming, so here it is :

I've implemented a variation of the MiniMax algorithm (returning moves instead of min/max values).

I've also tried setting it up as an alpha-beta, although it ended up a complete failure.

So, here's my MiniMax code :

Move* Board::miniMax(int depth)
{
    return this->maxMove(1, depth);
}

Move* Board::maxMove(int ply, int depth)
{
    vector<Move*> moves = this->possibleMoves();
    int movesSize = moves.size();

    Move* maxMove = new Move(MINUS_INF);

    for (int i=0; i<movesSize; i++)
    {
        Move* move = moves[i];
        HASHMAKE(move,this);

        move->value = (ply<depth) ? (this->minMove(ply+1, depth))->value 
                                  : this->eval();

        maxMove = MAXMOVE(maxMove,move);

        UNHASHMAKE(move,this);
    }

    return maxMove;
}

Move* Board::minMove(int ply, int depth)
{
    vector<Move*> moves = this->possibleMoves();
    int movesSize = moves.size();

    Move* minMove = new Move(PLUS_INF);

    for (int i=0; i<movesSize; i++)
    {
        Move* move = moves[i];
        HASHMAKE(move,this);

        move->value = (ply<depth) ? (this->maxMove(ply+1, depth))->value 
                                  : this->eval();

        minMove = MINMOVE(minMove,move);

        UNHASHMAKE(move,this);
    }

    return minMove;
}

Any ideas one how the above thing could be adjusted so that it's an Alpha-Beta search?

---

And here's my attempt at Alpha-Beta conversion (which fails miserably) :

Move* Board::alphaBeta(int depth)
{
    return this->alphaMax(1,depth,MINUS_INF,PLUS_INF);
}

Move* Board::alphaMax(int ply, int depth, int a, int b)
{
    vector<Move*> moves = this->possibleMoves();
    int movesSize = moves.size();

    Move* maxMove = new Move(MINUS_INF);

    for (int i=0; i<movesSize; i++)
    {
        Move* move = moves[i];
        HASHMAKE(move,this);

        move->value = (ply<depth) ? (this->alphaMin(ply+1, depth,a,b))->value 
                                  : this->eval();

        maxMove = MAXMOVE(maxMove,move);
        if (maxMove->value>=b) return maxMove;
        a = MAXVAL(a,maxMove->value);

        UNHASHMAKE(move,this);
    }

    return maxMove;
}

Move* Board::alphaMin(int ply, int depth, int a, int b)
{
    vector<Move*> moves = this->possibleMoves();
    int movesSize = moves.size();

    Move* minMove = new Move(PLUS_INF);

    for (int i=0; i<movesSize; i++)
    {
        Move* move = moves[i];
        HASHMAKE(move,this);

        move->value = (ply<depth) ? (this->alphaMax(ply+1, depth,a,b))->value 
                                  : this->eval();

        minMove = MINMOVE(minMove,move);
        if (minMove->value<=a) return minMove;
        b = MINVAL(b,minMove->value);

        UNHASHMAKE(move,this);
    }

    return minMove;
}

---

HINTS (to avoid any misunderstanding) :

the this->eval() function returns a score from player A's perspective. Eg a +100 score means the position is in favour of player A, while a -100 score means the position is in favour of player B.

MINUS_INF and PLUS_INF have been defined as some arbitrarily small and big values, respectively.

This not anything like a homework or anything (if it was I'd most likely never have any interest to play with such stuff... lol)

Move is a simple class containing details about the move, as well as it's respective value (as assigned by the eval function.

HASHMAKE and UNHASHMAKE are just 2 move-(un)making and move-(un)hashing macros, that shouldn't make much difference.

MAXMOVE is defined like this : #define MAXMOVE(A,B) (((A)->value>=(B)->value)?(A):(B))

MINMOVE is defined like this : #define MINMOVE(A,B) (((A)->value<=(B)->value)?(A):(B))

---

Not sure if this is it, but I think in alphaMin

if (minMove->value<=a) return minMove;
b = MINVAL(b,minMove->value);
UNHASHMAKE(move,this);

should be

UNHASHMAKE(move,this);
if (minMove->value<=a) return minMove;
b = MINVAL(b,minMove->value);

and also a similar change in alphaMax .

链接地址: http://www.djcxy.com/p/9632.html

上一篇: 如何使用negamax算法

下一篇: 问题与MiniMax到Alpha