/// <summary>
/// public wrapper for recursion
/// try each possible move, find the one with the best score
/// </summary>
/// <param name="lookahead">how far to look ahead</param>
/// <param name="playerX">is this for player x</param>
/// <returns>The data on the best move location and score</returns>
public MinimaxResult DoMinimax(int lookahead, bool playerX)
{
// set up inital state
DateTime startTime = DateTime.Now;
if (lookahead < 1)
{
throw new Exception("Invalid lookahead of " + lookahead);
}
this.debugDataItems.Clear();
Occupied player = playerX.ToPlayer();
int alpha = MoveScoreConverter.ConvertWin(player.Opponent(), 0);
int beta = MoveScoreConverter.ConvertWin(player, 0);
MinimaxResult bestMove = this.ScoreBoard(lookahead, this.ActualBoard, playerX, alpha, beta);
if (bestMove.Move != Location.Null)
{
GoodMoves.AddGoodMove(0, bestMove.Move);
}
DateTime endTime = DateTime.Now;
this.MoveTime = endTime - startTime;
return(bestMove);
}
public Minimax(HexBoard board, GoodMoves goodMoves, ICandidateMoves candidateMovesFinder)
{
this.actualBoard = board;
this.goodMoves = goodMoves;
this.candidateMovesFinder = candidateMovesFinder;
this.boardCache = new BoardCache(board.Size);
this.pathLengthFactory = new PathLengthAStarFactory();
}
/// <summary>
/// Initializes a new instance of the HexGame class, with a board size
/// </summary>
/// <param name="boardSize">size of the board</param>
/// ********
public HexGame(int boardSize)
{
this.board = new HexBoard(boardSize);
IPathLengthFactory pathLengthFactory = new PathLengthAStarFactory();
this.xPathLength = pathLengthFactory.CreatePathLength(this.board);
this.yPathLength = pathLengthFactory.CreatePathLength(this.board);
this.goodMoves = new GoodMoves();
this.goodMoves.DefaultGoodMoves(boardSize, 5);
}
/// <summary>
/// private recursive worker - does the minimax algorithm
/// </summary>
/// <param name="lookahead">the current ply, counts down to zero</param>
/// <param name="stateBoard">the current board</param>
/// <param name="isPlayerX">player X or player Y</param>
/// <param name="alpha">alpha value used in alpha-beta pruning</param>
/// <param name="beta">beta value used in alpha-beta pruning</param>
/// <returns>the score of the board and best move location</returns>
private MinimaxResult ScoreBoard(
int lookahead,
HexBoard stateBoard,
bool isPlayerX,
int alpha,
int beta)
{
this.CountBoards++;
MinimaxResult bestResult = null;
Location cutoffMove = Location.Null;
var possibleMoves = this.candidateMovesFinder.CandidateMoves(stateBoard, lookahead);
foreach (Location move in possibleMoves)
{
// end on null loc
if (move.IsNull())
{
break;
}
if (this.GenerateDebugData)
{
this.AddDebugDataItem(lookahead, move, isPlayerX, alpha, beta);
}
// make a speculative board, like the current, but with this cell played
HexBoard testBoard = this.boardCache.GetBoard();
testBoard.CopyStateFrom(stateBoard);
testBoard.PlayMove(move, isPlayerX);
MinimaxResult moveScore;
PathLengthBase staticAnalysis = this.pathLengthFactory.CreatePathLength(testBoard);
int situationScore = staticAnalysis.SituationScore();
if (lookahead <= 1)
{
// we have reached the limits of lookahead - return the situation score
moveScore = new MinimaxResult(situationScore);
}
else if (MoveScoreConverter.IsWin(situationScore))
{
// stop - someone has won
moveScore = new MinimaxResult(situationScore);
}
else
{
// recurse
moveScore = this.ScoreBoard(lookahead - 1, testBoard, !isPlayerX, beta, alpha);
moveScore.MoveWins();
}
moveScore.Move = move;
this.boardCache.Release(testBoard);
// higher scores are good for player x, lower scores for player y
if (bestResult == null || MoveScoreConverter.IsBetterFor(moveScore.Score, bestResult.Score, isPlayerX))
{
bestResult = new MinimaxResult(move, moveScore);
}
// do the alpha-beta pruning
alpha = CheckAlpha(alpha, moveScore.Score, isPlayerX);
if (IsAlphaBetaCutoff(isPlayerX, alpha, beta))
{
cutoffMove = move;
bestResult.Score = alpha;
break;
}
// end a-b pruning
}
if (bestResult != null)
{
GoodMoves.AddGoodMove(lookahead, bestResult.Move);
}
if (cutoffMove != Location.Null)
{
GoodMoves.AddGoodMove(lookahead, cutoffMove);
}
return(bestResult);
}
