I ran the new Board and search downloads 4/14. I was running it at 4 depth and when it put me in checkmate it did not say that it won.
put the computer against itself at depth 4 and 3 it ended up looping
when it could have put in check. White can checkmate black.
The new version 4/4 has these bugs:
1) does not recognize stalemate for either side.
2) ply 1 the black king repeatedly tried to move into check
3) ply 2 similiar situation not as frequent.
4) ply 2 computer had taken all my pieces and would not place me in checkmate.
5) Winboard allowed be to castle after I had already moved the king. Our engine recognized the error but winboard said that it was the computers turn it just froze on that indefinitly.
I have just three questions:
1) Would it be possible to make the search depth dynamic? (perhaps, when nodes per second are optimal and/or when there are fewer pieces on the board, the engine could search deeper)
2) Toward the end of the game, when our engine's movement becomes repetitive, could that be caused by the fact that the engine can not see far enough to devise a possible plan of action?
3) Does the engine consider the fact that a pawn will become a queen
when promoted before the fact?
Soon after the Chess engine does a pawn promotion the Rook will go into a loop and just move back and forth.
I have come across two situations when the game didn't recognize that it had me in check mate. Has anyone else had this problem?
See the screenshots of these situations.
I was running our chess engine against GNU chess today. When GNU put our queen in jeprody, our engine decided to capture a knight. On it's next move GNU easily captured our queen. It almost seemed as though our engine thought our queen was already lost and just tried to make the best of a bad situation. Hope this helps, the new version seems great! Although still very slow on depth 6, it finished a game at depth 5 in about ten minutes.
After the update today, it seems that this does not happen as often. It does still happen occasionally. The engine also seems to be a little slower. Fixed in 03/31/03 release by Jason.
I was running our engine against GNU chess on winboard and came across this problem: See the screenshot. Our engine is white and GNU is black. Our chess engine can't figure out what to do here... it's obvious that it should capture the bishop with a pawn, but it has been thinking about this move for over 40 minutes. Granted I was running it a 6-ply, but still... It made the previous moves in a reasonable time (2-6 minutes). This may be related to the problem where the engine tries to kill the opponent's king when it is put in check... I left the program running after I went to bed. When I checked it this morning I found that our engine had actually made 2 more moves :) So I guess it was working after all.
Also, playing a different game at 5 ply, when GNU put the engine in check, it tried to make the same move as it did for the previous move. The piece was not even there anymore. View this screenshot. Fixed in 03/29/03 release by Jason.