Check out Grant Acedrex, our featured variant for April, 2024.


[ Help | Earliest Comments | Latest Comments ]
[ List All Subjects of Discussion | Create New Subject of Discussion ]
[ List Earliest Comments Only For Pages | Games | Rated Pages | Rated Games | Subjects of Discussion ]

Single Comment

Tjatoer!. A chess variant written in Python, designed to stress test engines.[All Comments] [Add Comment or Rating]
💡📝Halfen Ludith wrote on Sat, Jun 26, 2021 03:48 PM UTC in reply to H. G. Muller from 03:23 PM:

I agree with the 5 fold repetition, it should stay the same as regular chess, although the reason I implemented the 100-move rule is that it acts as headroom for future engines to find longer mates, the game is already tedious to play for two humans but for engines (the reason I designed this game in the first place is to study machine learning), there some of the positions even in regular chess where mates of more than 50 moves are possible (I think we all know of the famous mate in 500 moves puzzles), and with a game this large, I bet there are more positions of that kind, so I don't want future engines to jump into conclusion to call for a draw when it sees a mate that is more than 50 moves away, and instead change their strategy to eliminate the last remaining piece of the losing side (hence the bare king rule), that being said, it is true that the counter should reset after a promotion

Regarding castling and en passant, well castling is obvious since you need to move a billion pieces to clear the way for the king, as for en passant, I figured even if the chance to do en passant occurs, it might be either too early in the game where there are adequate number of pieces to stop the passed pawn, or it's too late in the game the chance to do en passsant is astronomically low