Author: Amir Ban
Date: 16:17:59 07/23/00
Go up one level in this thread
On July 23, 2000 at 18:16:20, Peter McKenzie wrote:
>On July 23, 2000 at 13:23:50, Amir Ban wrote:
>
>>On July 22, 2000 at 17:58:06, Peter McKenzie wrote:
>>
>>>Well, the Deep Blue threads have started up again (!) which can be a bit boring
>>>although there are some interesting issues if we can ignore all the emmotional
>>>stuff. I'm personally not that interested in the technical stuff (NPS, ply
>>>depth etc) but am more interested in the quality of the moves (I think Amir said
>>>something similar).
>>>
>>>I think one of the most impressive computer chess games ever was game 1 from the
>>>1996 Deep Blue - Kasparov match:
>>>
>>>1. e4 c5 2. c3 d5 3. exd5 Qxd5 4. d4 Nf6 5. Nf3 Bg4 6. Be2 e6 7. h3 Bh5 8.
>>>O-O Nc6 9. Be3 cxd4 10. cxd4 Bb4 11. a3 Ba5 12. Nc3 Qd6 13. Nb5 Qe7 14. Ne5
>>>Bxe2 15. Qxe2 O-O 16. Rac1 Rac8 17. Bg5 Bb6 18. Bxf6 gxf6 19. Nc4 Rfd8 20.
>>>Nxb6 axb6 21. Rfd1 f5 22. Qe3 Qf6 23. d5 Rxd5 24. Rxd5 exd5 25. b3 Kh8 26.
>>>Qxb6 Rg8 27. Qc5 d4 28. Nd6 f4 29. Nxb7 Ne5 30. Qd5 f3 31. g3 Nd3 32. Rc7
>>>Re8 33. Nd6 Re1+ 34. Kh2 Nxf2 35. Nxf7+ Kg7 36. Ng5+ Kh6 37. Rxh7+
>>>and Kasparov (black) resigned.
>>>
>>>A powerful display from DB, and amazing that Kasparov was able to come back and
>>>win the match after what must have been a shocking loss.
>>>
>>>Anyway, I am curious about how hard it is for current micro programs to find the
>>>moves of DB in the critical stages of the game. To determine this, I have
>>>prepared a small test suite containing the white to move positions from move 18
>>>through to move 30:
>>>
>>>2r2rk1/pp2qppp/1bn1pn2/1N2N1B1/3P4/P6P/1P2QPP1/2R2RK1 w - - bm Bxf6; id
>>>"db96_18";
>>>2r2rk1/pp2qp1p/1bn1pp2/1N2N3/3P4/P6P/1P2QPP1/2R2RK1 w - - bm Nc4; id "db96_19";
>>>2rr2k1/pp2qp1p/1bn1pp2/1N6/2NP4/P6P/1P2QPP1/2R2RK1 w - - bm Nxb6; id "db96_20";
>>>2rr2k1/1p2qp1p/1pn1pp2/1N6/3P4/P6P/1P2QPP1/2R2RK1 w - - bm Rfd1; id "db96_21";
>>>2rr2k1/1p2qp1p/1pn1p3/1N3p2/3P4/P6P/1P2QPP1/2RR2K1 w - - bm Qe3; id "db96_22";
>>>2rr2k1/1p3p1p/1pn1pq2/1N3p2/3P4/P3Q2P/1P3PP1/2RR2K1 w - - bm d5; id "db96_23";
>>>2r3k1/1p3p1p/1pn1pq2/1N1r1p2/8/P3Q2P/1P3PP1/2RR2K1 w - - bm Rxd5; id "db96_24";
>>>2r3k1/1p3p1p/1pn2q2/1N1p1p2/8/P3Q2P/1P3PP1/2R3K1 w - - bm b3; id "db96_25";
>>>2r4k/1p3p1p/1pn2q2/1N1p1p2/8/PP2Q2P/5PP1/2R3K1 w - - bm Qxb6; id "db96_26";
>>>6rk/1p3p1p/1Qn2q2/1N1p1p2/8/PP5P/5PP1/2R3K1 w - - bm Qc5; id "db96_27";
>>>6rk/1p3p1p/2n2q2/1NQ2p2/3p4/PP5P/5PP1/2R3K1 w - - bm Nd6; id "db96_28";
>>>6rk/1p3p1p/2nN1q2/2Q5/3p1p2/PP5P/5PP1/2R3K1 w - - bm Nxb7; id "db96_29";
>>>6rk/1N3p1p/5q2/3Qn3/3p1p2/PP5P/5PP1/2R3K1 b - - bm Qd5; id "db96_30";
>>>
>>>Some of these moves will be very easy to find, but some are quite tough I think
>>>(23.d5, 25.b3 probably). Perhaps there are also other good moves in these
>>>positions, it will be interesting to see what various programs come up with.
>>>
>>>I just ran a very quick test using LambChop and it got 6/13 correct. I will run
>>>it again using more thinking time and post results.
>>>
>>>Have fun,
>>>Peter
>>
>>It's assumed that Deep Blue played perfectly, and it doesn't look like it.
>>Kasaprov went wrong on 27... d4? Instead 27... Qg5 gets a draw (28. g3 f4 29. g4
>>{29. Nd6 Qh5 30. Qc3+ d4 31. Qc5 Qxh3 draw } f5 30. Qxd5 hxg4 31. Qxg5 Rxg5
>>draw). Since the position after white's move 27 is a draw, there's not much
>>significance to DB's moves before that. If anything, it blew a small advantage
>>with the suspicious 26. Qxb6?! and 27. Qc5?!
>
>Interesting, is this known analysis or something you and/or Junior discovered?
>I have my program running at home to see if it finds a draw with 27...Qg5.
>
This is analysis by Junior.
>If 26.Qxb6 and 27.Qc5 are mistakes, it will be a good test to see if any
>programs can find better moves. What are the best moves in these positions?
>
Obviously 27.Qe3. An alternative to 26.Qxb6 is 26.Qf4 (intending Nd6). I
evaluate all these around +0.20. It's a moot point because the position is about
equal. I prefer the alternatives because after 26. Qxb6 Rg8 27.Qe3 d4 it's hard
to see white winning, while after 27.Qc5 Qg5, it's a draw but it's white who is
lucky not to lose.
Amir
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.