Author: Robert Hyatt
Date: 17:39:05 07/14/02
Go up one level in this thread
On July 14, 2002 at 10:46:19, Mike S. wrote:
>On July 13, 2002 at 18:05:38, Robert Hyatt wrote:
>
>>(...) But the right circumstance
>>must be hit... ie the capture opens the h file. The computer has castled
>>and the opponent has not. There must be rooks and a queen still on the board.
>>
>>If all that is true, I doubt I would _ever_ make the capture unless I can
>>in the same variation force the trade of queens or get rid of the rooks...
>
>I'd be interested what you think about the common Lxh7+ sacrifice in the French
>opening. Sometimes I was able to win against programs with that (slow comps
>compared to now though), because they castled to the kingside, which surely was
>the major mistake in those positions.
>
This is a different problem, in general. IE Bxh7 is a direct attack on
the king, and generally results in Kxh7 followed by something like Ng5+
giving the king a choice, back to the first rank, or forward.
I think, in general, that a program should be able to defend against this
king of attack by search alone. The check should extend the search and several
consecutive moves are very forcing. This kind of thing can work at very fast
time controls, since the program might not be able to see the consequences of
things early enough to stay out of trouble. But at longer (deeper) searches,
it ought to be able to avoid difficulty. The trojan horse is a different
animal. The computer makes the key move, and none of the early moves are
posing any significant threat that stands out due to checks, captures or
any sort of warning... I have seen players go to great lengths to disguise
this so that programs will fall for it...
>Note, that in both example games below, the white bishop was not yet on d3 when
>Black castled.
>
>And: In the popular set of Nunn Opening Positions, no. 3 is about the Lxh7+ sac,
>where Black should avoid the castling move too. Ny impression is though, that in
>many test games from Nunn matches, the result in that position often did not
>depend on if Black castled or not (or if White played Lxh7+ or not then)... but
>this would have to be researched further. I don't know if it's a forced win in
>that position, after ...O-O.
>
>[D]r3k2r/p1qbnppp/1pn1p3/2ppP3/P2P4/2PB1N2/2P2PPP/R1BQ1RK1 b kq - 0 11
>No. 3 from the Nunn Openings.
>
>Is this a similar danger for programs like the trojan sac (a little less
>dangerous probably), and does it require special code to avoid the dangerous
>castling? Or do you think, "normal" calculation and king safety evaluation
>should be enough nowadays, to prevent against that Lxh7+ sacrifice?
I would hope that a good search on decent hardware would avoid this by
direct means... unlike the trojan which is a longer-term problem..
>
>Thanks,
>Mike Scheidl
>
>P.S. The games:
>
>[Event "60/15:00 P133"]
>[Site "?"]
>[Date "1998.??.??"]
>[Round "?"]
>[White "Scheidl, M."]
>[Black "Comet_A.75"]
>[Result "1-0"]
>[ECO "C19"]
>
>1. e4 e6 2. d4 d5 3. Nc3 Bb4 4. e5 c5 5. a3 Bxc3+ 6. bxc3 Ne7 7. Nf3 Bd7 8. a4
>Qa5 9. Bd2 Nbc6 {end of book} 10. h4 cxd4 11. cxd4 Qc7 12. Rb1 O-O 13. Bd3 Rfc8
>14. Bxh7+ Kxh7 15. Ng5+ Kg8 (15... Kg6 16. h5+ Kh6 17. Nxe6+) 16. Qh5 Ng6 17.
>Qh7+ Kf8 18. Rh3 b6 19. Rf3 Ngxe5 (19... Be8 $2 20. Nxe6+) 20. dxe5 Qxe5+ 21.
>Kf1 Nd8 22. Bb4+ Rc5 23. Bxc5+ bxc5 24. Re1 (24. Nxf7 $1 {Fritz5} 24... Nxf7
>25. Qh8+ Ke7 26. Qxa8) 24... Qb8 25. Qh8+ Ke7 26. Qxg7 Be8 27. h5 Qc8 28. h6
>Qa6+ 1-0
>
>[Event "60/30:00 P133"]
>[Site "?"]
>[Date "1998.??.??"]
>[Round "?"]
>[White "Scheidl, M."]
>[Black "Diogenes 2.30"]
>[Result "1-0"]
>[ECO "C19"]
>
>1. e4 e6 2. d4 d5 3. Nc3 Bb4 4. e5 c5 5. a3 Bxc3+ 6. bxc3 Ne7 7. Nf3 Qa5 8. Bd2
>Nbc6 9. h4 O-O 10. Bd3 c4 (10... h6 {Fritz5} 11. Ng5 c4 12. Bh7+ Kh8 13. Qh5
>Qa4 14. Rh3 Nd8 15. Rf3 Qd7 16. Rf6 $1 Ng8 17. Bxg8 Kxg8 18. Ne4 dxe4 19. Bxh6
>g6 20. Qg5 Re8 21. h5 Qe7 22. hxg6 fxg6 23. Rxg6+ Kh8 24. Bg7+ Kg8 25. Bf6+ Kf7
>26. Rg7+ Kf8 27. Bxe7+ Rxe7 28. Qxe7#) 11. Bxh7+ Kxh7 12. Ng5+ Kg8 13. Qh5 Rd8
>14. Qxf7+ Kh8 15. Rh3 (15. h5 $1 Nf5 16. h6 Nxh6 17. Rxh6+ gxh6 18. Qh7#) 15...
>Nf5 16. g4 Nxe5 17. dxe5 Qb6 18. gxf5 exf5 19. Qh5+ Qh6 20. Nf7+ Kg8 21. Nxh6+
>gxh6 22. Bxh6 Kh7 23. Rg3 Rd6 24. Rg7+ Kh8 25. Qe8# 1-0
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.