Author: George Sobala
Date: 14:34:29 01/11/04
Go up one level in this thread
On January 11, 2004 at 11:27:40, Bob Durrett wrote: >On January 11, 2004 at 05:43:04, Jouni Uski wrote: > >>[D]r1r3k1/4bp1p/1qp1b1p1/1p1n4/3PB3/P5N1/1BQ2PPP/R3R1K1 w - - 0 21 >> >>In this position Ruffian plays almost instantly Bxg6. Is this real sacrifice? >>Is it correct? Actually Ruffian lost this game against Fritz... >> >>Jouni >> >>PS: I hope, that next Ruffian update gives the missing 70 points from promised >>100 over 1.0.1 :-) > >Chess engines have long been able to find tactical combinations. Combinations >often begin with a pseudosacrifice such as Bxg6. > >As noted in other threads, the so-called "true" sacrifice is merely an EXCHANGE >of an advantage in material for an equivalent [or better] advantage in position. > It seems to me that engines equipped with sophisticated position evaluation >code [and compatible searching] should be able to recognize the equivalence of >positional advantages to material advantages. The programmer, however, must >"allow" his/her engine to play positional chess, including positional >sacrifices. Too much "material-oriented" code may make positional chess >impossible, I suspect. The chess programmer who plays chess at the 1500 level >may not comprehend positional chess and positional sacrifices and as a result >inadvertently put code into his/her engine which precludes the engine's playing >positional chess. > >Chess engines will probably never play psychological sacrifices but may someday >play sacrifices based on general strategic considerations. Hopefully. > >Bob D. I beg to differ. I have tweaked the personality of Deep Sjeng to create variants which place _huge_ and disproportionate emphasis on positional features such as piece position, space and king attack. They do not play "correct" chess, in that materialistic engines usually beat them, but they love tossing material for what they fondly imagine is a positional advantage. Against human players up to FM / low IM strength they can be formidable opponents, often getting into objectively lost but subjectively very scarey and pressurised positions. I consider these to be psychological sacrifices! The humans usually crack under the strain, blunder and lose. As a little example of how such an engine can even beat a comp have a look at this game: redshift (my variant) playing ajop, one of the stronger computers on ICC, in one of this week's computer tournaments. Ajop was a "sensible" engine (ChessTiger-15) on a 50% faster machine. [Event ""] [Site "chessclub.com"] [Date "2004.1.6"] [Round ""] [White "ajop"] [Black "redshift"] [Result "0-1"] [Eco ""] [Annotator ""] [Source ""] 1.d4 d5 2.c4 c6 3.Nf3 Nf6 4.Nc3 g6 5.cxd5 cxd5 6.Bf4 Bg7 7.e3 Nc6 8.h3 Nh5 9.Be5 f6 10.Bh2 f5 11.Ne5 e6 12.Nxc6 bxc6 13.Be2 Qh4 14.Qa4 Bd7 15.O-O O-O 16.Qb4 Rf7 17.Na4 f4 18.Bxh5 Qxh5 19.Bxf4 e5 20.dxe5 Bxh3 21.gxh3 Qxh3 22.Bg3 g5 23.Rfc1 Rf3 24.Qd2 Rc8 25.Rc2 h5 26.Qd4 h4 27.Bh2 Rcf8 28.Nc3 g4 29.Nxd5 g3 30.Ne7+ Kf7 31.fxg3 hxg3 32.Qd6 Bh6 33.e6+ Kg7 34.Qe5+ Kh7 35.Qe4+ Kh8 36.Ng6+ Kg7 37.Nf4 Bxf4 38.exf4 R8xf4 39.Qe5+ Kh7 40.Qc7+ Kh6 41.Qxf4+ Rxf4 42.Rg2 Rg4 43.Bxg3 Rxg3 0-1 Position after 20. dxe5 [D] r5k1/p2b1rbp/2p3p1/3pP2q/NQ3B2/4P2P/PP3PP1/R4RK1 b - - 0 1 No other engine would consider Bxh3 So this time, I even managed to hypnotise a Tiger. To add insult to injury, redshift went on to beat wackyraces in the next game, (holder of the highest ever rating on ICC, a comp which loses a game about twice a month) gaving got into a position where it was two pawns down and even _it_ thought it was losing! Remember - to win in chess it is not necessary to play the best moves, it is enough to make your opponent play bad ones.
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.