Author: Eelco de Groot
Date: 03:12:56 01/19/06
The game:
[Event "Corus Chess 2006"]
[Site "Wijk aan Zee"]
[Date "2006"]
[Round "Round 2"]
[White "Adams"]
[Black "Topalov"]
[Result "1-0"]
1. e4 c5 2. Nf3 d6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 a6 6. Be2 e6 7. O-O Be7 8. a4
Nc6 9. Be3
O-O 10. f4 Qc7 11. Kh1 Re8 12. Bf3
{[D]r1b1r1k1/1pq1bppp/p1nppn2/8/P2NPP2/2N1BB2/1PP3PP/R2Q1R1K b - -}
Bf8 13. Qd2 Na5 14. b3 Rb8 15. Rad1 Nc6 16. Bf2 Nd7
17. Bg3 Nxd4 18. Qxd4 b5 19. axb5 axb5 20. b4 g6 21. e5 d5 22. f5 gxf5 23. Nxd5
Qc4 24. Qd2
h6 25. h3 exd5 26. Bxd5 Qxb4 27. c3 Qc5 28. Rxf5 Re6 29. Rxf7 Nb6 30. Rdf1 Nxd5
31. Rxf8+
Qxf8 32. Rxf8+ Kxf8 33. Qxd5 Ke8 34. Bh4 Bd7 35. Bf6 b4 36. Qe4 Bc8 37. cxb4 Rb7
38. Qg6+
Kd7 39. Qxh6 Kc7 40. Qf4 Kb8 41. h4 Rc7 42. h5 1-0
Depth of 22 plies seems to be a practical limit for which to do this analysis
with Fruittype programs used, to which I also consider Glaurung, at least in
some of its design philosophy; the K.I.S.S. priciple in chess programming.
I still think there is reason to expect that the 12..Bf8 variation in the
Sicilian Scheveningen may face some revisions. To recap from my first post;
computers consider 13..Bf8-e7 often the strongest move, but this Bishop had just
moved back from this square a move earlier!
http://www.talkchess.com/forums/1/message.html?480199
Topalov played 13. Na5 in the game but the Knight went back to c6 a few moves
later. This was criticized by Adams afterwards, in my investigation I wanted to
see if opening problems could have arisen already earlier.
Mickey's comments on Coruschess:
http://www.coruschess.com/report.php?year=2006&report=2
"Topalov against Adams was a Najdorf-turned-Scheveningen Sicilian, repeating
their opening from San Luis. Veselin was less successful this time though, since
Adams developed a strong initiative. Mickey was critical of 15..Nc6?!
inconsistently retreating the recently moved Na5. He also called 20..g6?! A
‘suspicious move’(for better or worse 20..Qc4 had to have been tried). After the
game Topalov said he should have tried 21..Rb6!? Instead of 21..d5, which was
hit with the devilish 22.f5!"
The position after 13.Qd2:
[D]r1b1rbk1/1pq2ppp/p1nppn2/8/P2NPP2/2N1BB2/1PPQ2PP/R4R1K b - -
Pro Deo 1.1 Vulcan {9k17pr} prefers 13..e5. I wished I could have analysed
deeper but the 19.00 ply result for 13..e5, still first choice at 19.00, simply
took too long
r1b1rbk1/1pq2ppp/p1nppn2/8/P2NPP2/2N1BB2/1PPQ2PP/R4R1K b - -
00:00:00 1.00 -1.08 13..Nxd4 14.Bxd4
00:00:00 1.02 -0.75 13..Be7
00:00:00 1.03 -0.61 13..Bd7
00:00:00 2.00 -0.71 13..Bd7 14.Be2
00:00:00 2.07 -0.55 13..e5 14.Nf5 Bxf5 15.exf5
00:00:00 3.00 -0.38 13..e5 14.Nf5 d5 15.Nxd5
00:00:00 4.00 -0.57 13..e5 14.fxe5 Nxe5 15.Be2 Bg4
00:00:00 4.02 -0.45 13..Nxd4 14.Bxd4 e5 15.Be3 Be6
00:00:00 5.00 -0.36 13..Nxd4 14.Bxd4 e5 15.Be3 exf4 16.Bxf4 Qc5
00:00:00 6.00 -0.44 13..Nxd4 14.Bxd4 e5 15.Be3 exf4 16.Bd4 Be7
17.Qxf4
00:00:00 6.01 -0.38 13..e5 14.Nde2 Be6 15.Nd5 Bxd5 16.exd5
00:00:00 7.00 -0.30 13..e5 14.Nde2 Be7 15.f5 Bd7 16.Nd5 Nxd5 17.Qxd5
00:00:00 8.00 -0.34 13..e5 14.Nde2 Bd7 15.Rad1 Be7 16.Nd5 Nxd5
17.Qxd5
00:00:00 9.00 -0.37 13..e5 14.Nde2 Bd7 15.Nd5 Nxd5 16.exd5 e4
17.dxc6 exf3 18.cxd7
00:00:02 10.00 -0.24 13..e5 14.Nde2 Bd7 15.Rad1 Be7 16.Ng3 exf4
17.Bxf4 Qb6 18.Nf5 Qxb2
00:00:05 11.00 -0.24 13..e5 14.Nb3 exf4 15.Bxf4 Ne5 16.Be2 Be7
17.Rad1 Be6 18.Nd4 Bc4
00:00:12 12.00 -0.23 13..e5 14.Nb3 exf4 15.Bxf4 Ne5 16.Be2 Be6
17.Rad1 Bc4 18.Bxe5 Rxe5 19.Bxc4 Qxc4 20.Qd3
00:00:37 13.00 -0.24 13..e5 14.Nb3 Bd7 15.Be2 exf4 16.Rxf4 Be7
17.Nd4 Ne5 18.Nf5 Rac8
00:02:07 14.00 -0.22 13..e5 14.Nb3 Be6 15.Rad1 Nb4 16.Be2 Rad8
17.a5 Qc6 18.Bb6 Rc8
00:05:36 15.00 -0.25 13..e5 14.Nde2 Be7
00:18:16 16.00 -0.22 13..e5 14.Nde2 Be7
01:12:00 17.00 -0.27 13..e5 14.Nde2 Bd7 15.Ng3 exf4
01:21:54 17.01 -0.26 13..Nxd4 14.Bxd4 e5 15.Be3 Be7
03:33:33 18.00 -0.29 13..Nxd4 14.Bxd4 e5 15.Bf2 Be7 16.fxe5 dxe5
17.Rfe1
03:34:09 18.01 -0.27 13..e5 14.Nde2 Bd7
All results on 2009 MHz Athlon 200 Mb HT but most of the time two programs had
to share CPU time so running only at 50% of max speed.
Glaurung's result after 22 plies, two best moves, Dutch notation D=Q, P=N, T=R,
L=B:
Glaurung by Tord Romstad, see http://www.glaurungchess.com
Engine: Glaurung 1.0 (200 MB)
gemaakt door Tord Romstad
r1b1rbk1/1pq2ppp/p1nppn2/8/P2NPP2/2N1BB2/1PPQ2PP/R4R1K b - -
+---+---+---+---+---+---+---+---+
8 |*T*|:::|*L*|:::|*T*|*L*|*K*|:::
+---+---+---+---+---+---+---+---+
7 |:::|*p*|*D*| |:::|*p*|*p*|*p*
+---+---+---+---+---+---+---+---+
6 |*p*|:::|*P*|*p*|*p*|*P*| |:::
+---+---+---+---+---+---+---+---+
5 |:::| |:::| |:::| |:::|
+---+---+---+---+---+---+---+---+
4 | p |:::| | P | p | p | |:::
+---+---+---+---+---+---+---+---+
3 |:::| | P | | L | L |:::|
+---+---+---+---+---+---+---+---+
2 | | p | p | D | |:::| p | p
+---+---+---+---+---+---+---+---+
1 | T | |:::| |:::| T |:::| K
+---+---+---+---+---+---+---+---+
a b c d e f g h
Glaurung 1.0
-------------------------------------------------------------------------------
21 704:38 -0.20 13...Le7 14.Tad1 Pxd4 15.Dxd4 e5 16.Dd2 Le6 17.fxe5
dxe5 18.Pd5 Dc6 19.b3 Pg4 20.Pxe7+ Txe7 21.Lg5 Td7 22.De2 Txd1 23.Txd1 Pf6
24.Lxf6 gxf6 (8.754.554.527) 207
21 704:38 -0.29 13...e5 14.Pde2 Le7 15.f5 Pb4 16.Pg3 a5 17.Tfc1 Td8
18.Pb5 Dc6 19.Td1 Dxc2 20.Lb6 Ld7 21.Lxd8 Txd8 22.Pxd6 Lxa4 23.Dxc2 Lxc2
(8.754.554.527) 207
-------------------------------------------------------------------------------
Gambit Fruit was the third program to tackle this position and went even deeper
in time; 45 hours at 50 % CPU time, three best moves.
-------------------------------------------------------------------------------
15:01:44.8 -0,49 20 3262722585 Rb8 Nb3 b6 Rfe1 Be7 g4 Nd7 g5 Nc5 Rad1 Bd7 Bh5
Nb4 a5 Nxb3 axb6 Qxb6 cxb3 Qa5 Qd4
15:01:44.8 -0,55 20 3682259732 Be7 Qf2 e5 Nf5 Bxf5 exf5 Rac8 g4 e4 Bb6 Qd7 Nxe4
Nxe4 Bxe4 Bf6 Rae1 Bxb2 Qd2 Re7 Re3 Rce8 Rfe1 Kh8
15:01:44.8 -0,59 20 2040438825 e5 Nde2 Qa5 h3 Bd7 f5 Rad8 g4 d5 exd5 e4 Bg2 Nb4
Nd4 Bc5 Rad1 b6 Bg5 Bxd4 Qxd4 Nxc2
-------------------------------------------------------------------------------
45:38:21.4 -0,50 21 3997906813 Be7 Qf2 e5 Nf5 Bxf5 exf5 Rac8 g4 e4 Bb6 Qd7 Nxe4
Nxe4 Bxe4 Bf6 Rae1 Bxb2 Qd2 Nd8 Rf3 h5 gxh5 Qxa4 Rfe3
45:38:21.4 -0,53 21 4251500518 e5 Nde2 Nb4 a5 d5 fxe5 Nxe4 Nxe4 dxe4 Bxe4 Qxe5
Nc3 Bg4 Bb6 Bd6 g3 Qe7 Rae1 Qd7 Nd5 Nxd5 Qxd5 Bb4 Qxd7 Bxd7 Bxb7 Bxe1 Bxa8
45:38:21.4 -0,56 21 903855553 Rb8 Bf2 Kh8 Nb3 d5 Rad1 Bb4 exd5 exd5 Nxd5 Nxd5
Qxd5 Be6 Qh5 g6 Qg5 Be7 Qg3 Bf6 Bxc6 Qxc6 Bd4 Bxd4 Nxd4 Qxa4 Qg5 Rbd8 Qe5+ Kg8
Nxe6 Rxd1 Rxd1
-------------------------------------------------------------------------------
Analysis of the position one move earlier is still going because I wanted to do
the first position to at least 22 plies.
At this point in the Scheveningen many moves have been tried in practice
although 12..Bf8 seems to be the main choice. I haven't got an extended database
but this are the statistics from the Sandro's big Shredder 9 book:
Move pref. chance No.games Result Year
12..Bf8 a 86.0% 86 50.5% 2000
12..Bd7 c 11.5% 94 41.4% 1994
12..Rd8 d 1.8% 253 39.5% 1993
12..Nd7 d 0.5% 2 75.0% 1995
12..Na5 e 0.0% 76 53.2% 1994
12..Nxd4 e 0.0% 31 41.9% 1992
12..e5 f 0.0% 2 0.0% 1990
So far I have a Glaurung analysis-result for this position, two best moves:
[D]r1b1r1k1/1pq1bppp/p1nppn2/8/P2NPP2/2N1BB2/1PP3PP/R2Q1R1K b - -
Engine: Glaurung 1.0 (200 MB)
gemaakt door Tord Romstad
-------------------------------------------------------------------------------
21 338:35 -0.07 12...Pa5 13.Pde2 Pc4 14.Lc1 d5 15.exd5 Td8 16.f5 exf5 17.Dd4
Dc5 18.a5 Dxd4 19.Pxd4 Pe5 20.Le2 Pxd5 21.Pxf5 Lb4 22.Pxd5 Txd5 (4.266.440.004)
210
21 338:35 -0.10 12...e5 13.Pf5 Lxf5 14.exf5 e4 15.Pxe4 Pxe4 16.Lxe4 Lf6
17.Df3 De7 18.Lxc6 bxc6 19.Tae1 De4 20.Dxe4 Txe4 21.b3 Tb8 22.Tf3 d5
(4.266.440.004) 210
-------------------------------------------------------------------------------
22 1163:06 -0.06 12...Pa5 13.Pde2 Pc4 14.Lc1 d5
15.exd5 Td8 16.f5 exf5 17.Dd4 Dc5
18.Dxc5 Lxc5 19.Pg3 Pe3 20.Lxe3 Lxe3
21.Ph5 Pg4 22.Lxg4 fxg4 23.a5 (14.697.005.115) 210
22 1163:06 -0.14 12...e5 13.Pf5 Lxf5 14.exf5 e4
15.Pxe4 Pxe4 16.Lxe4 Lf6 17.Df3 De7
18.Lxc6 bxc6 19.Tae1 De4 20.Dxe4 Txe4
21.b3 Tae8 22.Tf3 d5 23.Tg3 (14.697.005.115) 210
-------------------------------------------------------------------------------
Glaurung 1.0 analysing at the moment 12..Rd8 (5/35) after 3280 minutes non-stop
analysis, I'm afraid completing the whole 22 ply iteration will take too long.
Third move considered at 22 plies was 12..Rb8 which took a long time but did not
get a better eval than the -0.14 for 12...e5.
So all in all, some computer conclusions: two move choices 12..e5 and 12..Rb8 at
move twelve seem similar, to the computer at least, to the same moves 13..e5 and
13..Rb8 considered after 12..Bf8 13.Qd2. After 12..Bf8 13.Qd2 the position seems
to be deteriorated for Black (compare evals)
At move thirteen the Knight manoever to a5 is not considered by Pro Deo Vulcan,
Glaurung and Gambit Fruit any more among the one, two cf. three best, only at
move twelve (Glaurung's analysis, it will be interesting to see if Gambit Fruit
or Vulcan concurs) and there so far it is first choice for Glaurung, 12..Bf8 is
not considered good by GF and Glaurung and simply placed back to e7 at move
thirteen.
12..Na5 seems to be the computer favourite so far after 50 hour analysis by
Glaurung. I could have played test games and analysis of deeper positions but
that was not the goal right now, just to see what comes out of an extended
overnight analysis and to see if the computer can come near what you might call
GMP-quality, GrandMasterPreparation-quality. Not quite yet I think, at least not
unaided within 8-24 hours, and not with one single program.
The computer maybe does not understand the ideas and same undoubtedly applies at
least doubly or tripley to me. But the programs I hope are not totally on the
wrong path here after these long calculations!
Eelco
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.