4-player-chess: computerprogram

[tsoj]

Since chess.com released a 4-player-chess mode it got quite popular I thought about how to port my existing chess engine to this special chess mode.
However, the longer I am thinking about that I find more differences.
First of all, should I still use bitboards for the board representation? I do not have alot of experience with bitboards, i do not really know if the advantages of bitboards apply if I use 3 bitboards for every piece instead of one. The simple alternative would be maybe the equivalent to a 12x10 board, a 16x16 board, however, that would waste a lot of memory...

I would say the min-max stays almost the same, however, I believe for 4-player-chess I can't use a single value for all players, maybe a structure of four values would be enough, one evaluation-value for every player.

The heuristic evaluation is probably the biggest problem. I believe I am not wrong if I say that I can't just sum up all pieces to get a rough winning probability. The evaluation function has to allow for strategic patterns. The best example is that in normal chess it is mostly ok to trade pieces, however, trading a queen in 4-player-chess is just horrible.
And then there is this point-system needs to be implemented.
I think there a lot of ways to do the evaluation function and a lot of them would be very bad and I fear that my ideas are one of them.

Is someone already working on a 4-player-chess program and has some tips how to do the evaluation right?

By the way, is it possible to have a 4-player-chess extension for the UCI protocol? Am I right if I say that we just need a fen- and move-notation for the 4-player-chess board?

[hgm]

Move notation in 4-player chess need not be different from that in regular chess: just the coordinates of from- and to-square, and possibly a promotion suffix. As for position notation: You might consider DEMN, which in this case is just 4 FEN board parts, one for each player (only listing his own pieces), all using capitals only, separated by '#' signs.

I think Daniel Shawul has worked on 4-player Chess.

[Harald]

My guess for the evaluation is:
First do 4 individual static evaluations, one for each player and his pieces.
Then each player who has the right to move tries to maximize the term
his_evaluation - max(the other evaluations)