And stockfish algorithm is quite simple, even if it was largely improved by having a more precise position evaluation, the only idea is to maximize player gain on each move, at every depths.Ī 4 player stockfish doesnt have to be any different, having a simple position evaluation function working would be enough to compute at a decent depth, maximizing each player gain on their own move.Įx: red and yellow can double checkmate green with double queen attack (simple example) red though process: if i take this pawn with my queen to check green, blue has no good move and yellow can trade his queen for a checkmate on green, if yellow doesn’t, i lose my queen and we will have a weak 2v2, if yellow does, it s 3 player and he is in a better position than blue even with a queen missing. You are wrong about how ia process, a simple ia like stockfish, even if it doesn’t learn, can make decent move (decent enough to beat weak human beings). That will give an equal evaluation on the starting position, and it would be using a minimaxminimax algorithm. The chess engine could do this through the form of arrows, where it draws an arrow for its teammate to help them decide what move to make. In Teams, the chess engine will assume that its teammate will always play the moves to maximize the score of itself. So it would be using a minimaxmaxmax algorithm. That would mean that in FFA, the initial evaluation of the start position would definitely be a huge negative value, since it thinks that all of the other players will all be attacking it to death, but that will prevent any cases of broken alliances causing unfavorable positions. This is because if it decides to try to make an alliance and that alliance is broken, it would be in a worst position if it decided not to make the alliance in the first place. In FFA, a 4 player chess engine should always assume that the opponent will play moves that minimize the chances of itself from winning. I think 4P Chess engines should work like this: This is because if the players play absolutely perfectly against the chess engine, there is no way that the chess engine can win, since it is 3v1. That actually makes the initial evaluation of the beginning position probably around -7 or maybe -10 in FFA. To me, I think a chess engine should assume that all players on the board want to do everything they can to checkmate the chess engine. That is why I think it actually isn't correct to make alliances in 4P from a chess engine perspective. That makes your ally have a much greater chance of ultimately winning the match because you decided to make that alliance.

You could make a sacrifice to allow your ally to checkmate them, but because you allowed your ally to checkmate them, they gain 20 points, while you gain no points (since the sacrificed piece was taken). Let's say you make an alliance with someone, and you are trying to checkmate another player with the help of that ally. The tricky part is when to actually make those alliances. All you have to make the AI do is assume that its ally will make moves that maximizes the chance of itself winning.

It is actually incredibly simple to create alliances.

3 way chess online vs computer