public IntPair GetAIMove()
{
// strategy 1: try all moves and choose the one with best AIscore
List <IntPair> possibleMoves = GetPossibleMoves();
// self-explanatory
if (possibleMoves.Count == 0)
{
throw new Exception("No moves for AI.");
}
else if (possibleMoves.Count == 1)
{
return(possibleMoves[0]);
}
List <AIMinimax> scoredMoves = new List <AIMinimax>();
// The score of the simulated move is calculated for the opponent.
// Therefore we want to minimize it.
// Randomize equal scores a bit.
foreach (IntPair move in possibleMoves)
{
Board sim = SimulateMove(move.i, move.j);
scoredMoves.Add(new AIMinimax {
pos = move, score = sim.AIScore(), rand = UnityEngine.Random.Range(0, 100)
});
}
// sort moves
scoredMoves.Sort((x, y) => (10000 * x.score + x.rand).CompareTo(10000 * y.score + y.rand));
// normalize
// this is just plain heuristics
// keep cBest best options
// the worst option gets score 1, better gets more
// just weigh best cBest options w.r.t. normalized weighths
int baseScore = scoredMoves[scoredMoves.Count - 1].score + 1;
int cBest = Math.Min(3, scoredMoves.Count);
scoredMoves = scoredMoves.GetRange(0, cBest);
int randPool = 0;
foreach (AIMinimax move in scoredMoves)
{
move.score = baseScore - move.score;
move.score = move.score * move.score; // square
randPool += move.score;
}
int rand = UnityEngine.Random.Range(0, randPool);
int m = 0;
while (scoredMoves[m].score < rand)
{
rand -= scoredMoves[m].score;
m++;
}
// TODO
string dbg = "";
foreach (AIMinimax move in scoredMoves)
{
dbg += " " + move.score;
}
Debug.Log("Normalized AIScores: " + dbg + " Chosen: " + scoredMoves[m].score);
return(scoredMoves[m].pos);
// int r = UnityEngine.Random.Range(0, possibleMoves.Count);
// return new IntPair(possibleMoves[r].i, possibleMoves[r].j);
}
