public int Evaluate(Board.Board board, int evaluatingPlayer)
{
int score = 0;
int winningPositionFor = board.CheckTerminalPosition();
int signum = evaluatingPlayer == 0 ? 1 : -1;
if (winningPositionFor == 0)
{
score += 1000;
}
if (winningPositionFor == 1)
{
score -= 1000;
}
score = score * signum;
score = score +
CountGoldCaptures(board) * GOLD_CAPTURES_WEIGHT * signum +
CountSilverCaptures(board) * SILVER_CAPTURES_WEIGHT * signum +
CalculateSouthBias(board) * SOUTH_BIAS_WEIGHT / board.width * signum +
CalculateFlagshipLiberties(board) * FLAGSHIP_LIBERTY_WEIGHT * signum;
return(score);
}
public int Evaluate(Board.Board board, int searchingPlayer)
{
int score = 0;
int winningPositionFor = board.CheckTerminalPosition();
if (winningPositionFor == 0)
{
score += 1000;
}
if (winningPositionFor == 1)
{
score -= 1000;
}
score = searchingPlayer == 0 ? score : (score * -1);
return(score);
}
private int AB(ref Board.Board board, int alpha, int beta, int depth) // maybe can implement minimal search window and deep pruning by giving alpha and beta as reference
{
if (depth == 0 || board.CheckTerminalPosition() >= 0)
{
return(evaluationHeuristic.Evaluate(board, board.activePlayer) + random.Next(-randomRange, randomRange));
}
LinkedList <(int, int)> moves = board.GetLegalMoves();
if (moves.Count == 0)
{
return(evaluationHeuristic.Evaluate(board, board.activePlayer) + random.Next(-randomRange, randomRange));
}
int score = int.MinValue / 2;
int value = score;
foreach ((int, int)move in moves)
{
board.Move(move);
if (board.remainingActions <= 1)
{
value = AB(ref board, Math.Max(alpha, score), beta, depth - 1);
}
else
{
value = -AB(ref board, -1 * beta, -1 * Math.Max(alpha, score), depth - 1);
}
if (value > score)
{
score = value;
}
board.Undo();
if (score > beta)
{
break;
}
}
return(score);
}
public int Evaluate(Board.Board board, int evaluatingPlayer)
{
int score = 0;
int winningPositionFor = board.CheckTerminalPosition();
int signum = evaluatingPlayer == 0 ? 1 : -1;
if (winningPositionFor == 0)
{
score += 1000;
}
if (winningPositionFor == 1)
{
score -= 1000;
}
score = score * signum;
score += (JudgeMaterialBalance(board) * MATERIAL_BALANCE_WEIGHT * signum);
return(score);
}
private int AB(ref Board.Board board, int alpha, int beta, int depth) // maybe can implement minimal search window and deep pruning by giving alpha and beta as reference
{
if (depth == 0)
{
//selective deepening in case of captures, might want to extend to other more or less forcing moves ("check", looming escpae)
// right now seems to slow down the calculations considerably, if tt doesnt help, might consider also to simplify it to only search recapturing sequences
if (board.captures.Count > 0 && board.captures.Last.Value.Item1 == (board.turnCounter - 1))
{
int selectScore = int.MinValue / 2;
int selectValue = selectScore;
LinkedList <(int, int)> forcingMoves = board.GetForcingMoves();
foreach ((int, int)move in forcingMoves)
{
board.Move(move);
if (board.remainingActions <= 1)
{
selectValue = AB(ref board, Math.Max(alpha, selectScore), beta, 0);
}
else
{
selectValue = -AB(ref board, -1 * beta, -1 * Math.Max(alpha, selectScore), 0);
}
if (selectValue > selectScore)
{
selectScore = selectValue;
}
board.Undo();
if (selectScore > beta)
{
break;
}
}
return(selectScore);
}
return(evaluationHeuristic.Evaluate(board, board.activePlayer) + random.Next(-randomRange, randomRange));
}
if (depth == 0 || board.CheckTerminalPosition() >= 0)
{
return(evaluationHeuristic.Evaluate(board, board.activePlayer) + random.Next(-randomRange, randomRange));
}
LinkedList <(int, int)> moves = board.GetLegalMoves();
if (moves.Count == 0)
{
return(evaluationHeuristic.Evaluate(board, board.activePlayer) + random.Next(-randomRange, randomRange));
}
int score = int.MinValue / 2;
int value = score;
foreach ((int, int)move in moves)
{
board.Move(move);
if (board.remainingActions <= 1)
{
value = AB(ref board, Math.Max(alpha, score), beta, depth - 1);
}
else
{
value = -AB(ref board, -1 * beta, -1 * Math.Max(alpha, score), depth - 1);
}
if (value > score)
{
score = value;
}
board.Undo();
if (score > beta)
{
break;
}
}
return(score);
}