Author: Mike S.
Date: 17:09:01 05/04/04
Go up one level in this thread
On May 04, 2004 at 19:04:51, Ed Trice wrote: >Does anyone have a concrete example of what position, or family of positions, >this code is encouraging the program to seek or avoid? From an old Machine vs. Man game; the comp was ChessMachine Gideon 3.1 (1993): [D]r3k2r/pp1n1pp1/2p1pn2/q6p/PbBP2b1/2N1PN1P/1P1BQPP1/R4RK1 w kq h6 0 12 The critical move is 12.hxg4(?) White didn't capture - yet, but... [D]2kr3r/pp1n1pp1/2p1pn2/q6p/PbBP2b1/2N1PN1P/1P1BQPP1/R2R2K1 w - - 0 13 13.hxg4(?) I put the ? i brackets, because here, I'm not sure if the sac has to be declined. But I guess - not having analysed very carefully - that White can escape here, when he returns the material in time. But there are indeed positions - of course, as these are what the code is made for - where the side which accepts such an offer is inavoidably lost after a deep tactical sequence (utilizing the open h file etc.). It is important to notice that antitrojan code will eventually decline such a sacrifice even when it was not correct or when the capture was playable. Recently, a "mysterious observation" was made in a game Diep-Shredder (CSVN Leiden 2004): [D]1k3rnr/ppp5/3bN3/3Q2Np/3p2n1/P2q3P/1P3PP1/R1B2RK1 w - - 0 22 Shredder had just played 21...Ne5-g4(?!) Diep didn't capture, but played 22.f4 (1-0 after 48 moves). But hxg4 was ok, ± 22. hxg4 hxg4 23. g3 Rf5 (23... Nf6 24. Qg2 $16) 24. Qg2 Ne7 25. Bf4 Bxf4 26. gxf4 $16 (Analysis by M.Gurevich). I guess both programs have antitrojan code and therefore didn't consider 22.hxg4 to be playable. This isn't a real trojan sac (bQ can't come in the h-file quickly, wQ guards h1 after g2-g3), but all the typical elements exist, so it probably results in "wrong diagnosis" :-) I've tested this with Crafty 17.14, where the trojan check was switchable still. As expected, it want's to capture immediatly without that code, but never with the trojan check on. Regards, M.Scheidl
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.