Geschrieben von:/Posted by: Ron Murawski at 15 July 2004 00:31:25:
Als Antwort auf:/In reply to: Horizon book needs some cleaning geschrieben von:/posted by: Olivier Deville at 13 July 2004 08:12:07:
Game Horizon 4.1b12 - Chess System Tal II from round 5 is currently running :
1.e4 c5 2.Nf3 d6 3.d4 cxd4 4.Nxd4 Nf6 5.Nc3 a6 6.Be3 e6 7.g4 e5 (first move out of book for CSTal) 8.Nf5 g6 9.Bg2?! from the book !
Actually, this piece sacrifice belongs to theory. But it seems to me this line is too speculative to be included in a tournament book for engines.
Olivier
Hi Olivier,
I have researched this move a little.
rnbqkb1r/1p3p1p/p2p1np1/4pN2/4P1P1/2N1B3/PPP2P1P/R2QKB1R w KQkq -
Horizon's move probabilities are:
g5: 85.71%
Bg2: 14.29%
Here's what my Crafty book says:
g5: 53%
Bg2: 46%
It is very difficult, perhaps impossible, to create a perfectly clean book based on pgn files. When you gather many thousands of games they will contain unsound, speculative moves for certain. Also, "won games" get lost by blunders and then the bad move sequences that led to these flawed wins get incorporated into engine books as the preferred moves. Perhaps I need to implement book learning to minimize these gaffs.
I think you are right that the move is too speculative for chess engines. Thanks for noticing this! Bg2 will be removed from the next book.
Best regards,
Ron
Horizon Home Page