Insanity... or Tal style?

[michiguel]

I have been tuning eval parameters in Gaviota with "apparent" success. First ~800 fast games suggest a possible increase in ~70 elo points from the version it played in CCT11. However, look at this position to see what Gaviota played.

[D]r4rk1/1bqnbppp/pp1ppn2/8/2PNP3/2N1B1P1/PP3PBP/2RQR1K1 w - - 5 13

First move after the setup position and it played Nd5?!?!?!? What?

I do not know where the heck that came from, and clearly, the tuning did not work in this position or... did it?
The funny thing is that Gaviota kept the game messy playing like Tal on LSD and won against both Fruit 1.0 (nullmove off to dumb it down a little) and EXchess. Against EXchess it was a draw at the end but Exchess blundered the endgame.
You would assume that you give away a piece against an engine and it is over, but not necessarily. I found both games funny to watch and borderline crazy.

Games were fast (40 moves/20 seconds), but not superfast. Engines reached at least 8 plies each move.

Miguel

[Event "Computer Chess Game"]
[Site "avefenix"]
[Date "2009.04.01"]
[Round "1"]
[White "Gaviota 0.59-test"]
[Black "Fruit 1.0"]
[Result "1-0"]
[TimeControl "40/20"]

1. c4 Nf6 2. Nf3 b6 3. g3 Bb7 4. Bg2 c5 5. O-O e6 6. Nc3 Be7 7. Re1 d6 8.
e4 a6 9. d4 cxd4 10. Nxd4 Qc7 11. Be3 Nbd7 12. Rc1 O-O 13. Nd5 exd5 14.
cxd5 Qd8 15. Nc6 Bxc6 16. dxc6 Nc5 17. b4 Ne6 18. c7 Nxc7 19. Bxb6 Nfe8 20.
Qc2 Qd7 21. e5 dxe5 22. Rxe5 Bd6 23. Rd1 Rb8 24. Ba5 Ne6 25. Red5 Qe7 26.
a3 Nd8 27. Bf1 Nb7 28. Qc6 Bc7 29. Rd7 Nxa5 30. Qxc7 Qxb4 31. axb4 Nxc7 32.
bxa5 Nb5 33. Bc4 Rbc8 34. Bxb5 axb5 35. a6 Ra8 36. a7 f6 37. Re7 Rf7 38.
Rdd7 Rxe7 39. Rxe7 h6 40. Kf1 Kh7 41. Ke2 Kg6 42. Kd3 f5 43. Kd4 Kf6 44.
Rc7 g6 45. Kc5 b4 46. Kb6 b3 47. Kb7 Rd8 48. a8=Q Rxa8 49. Kxa8 Ke5 50. Rb7
f4 51. Rxb3 fxg3 52. hxg3 Ke6 53. Rb6+ Kf5 54. g4+ Kg5 55. Kb7 h5 56. gxh5
gxh5 57. Kc6 h4 58. Kd5 Kf4 59. Rb4+ Kf3 60. Rxh4 Kxf2 61. Ke4 Kg3 62. Rf4
Kg2 63. Rf3 Kh2 64. Ke3 Kg2 65. Ke2 Kg1 66. Rg3+ Kh1 67. Kf3 Kh2 68. Kf2
Kh1 69. Rh3#
{Xboard adjudication: Checkmate} 1-0


[Event "Computer Chess Game"]
[Site "avefenix"]
[Date "2009.04.01"]
[Round "1"]
[White "Gaviota 0.59-test"]
[Black "EXchess_console"]
[Result "1-0"]
[TimeControl "40/20"]

1. c4 Nf6 2. Nf3 b6 3. g3 Bb7 4. Bg2 c5 5. O-O e6 6. Nc3 Be7 7. Re1 d6 8.
e4 a6 9. d4 cxd4 10. Nxd4 Qc7 11. Be3 Nbd7 12. Rc1 O-O 13. Nd5 exd5 14.
cxd5 Qb8 15. Nf5 Bd8 16. f3 b5 17. Bd4 a5 18. Qd3 Ba6 19. Qe3 Ne5 20. b3 b4
21. Bf1 Re8 22. Bxa6 Rxa6 23. f4 Ned7 24. Qd3 Qb7 25. e5 dxe5 26. fxe5 Nxd5
27. Nd6 Rxd6 28. exd6 Rxe1+ 29. Rxe1 N5f6 30. Qc4 Bb6 31. Re7 Bxd4+ 32.
Qxd4 Qb5 33. Qd2 Qc5+ 34. Kg2 Qc6+ 35. Kh3 Ne4 36. Re8+ Nf8 37. Rxf8+ Kxf8
38. d7 Ng5+ 39. Qxg5 Qxd7+ 40. Kg2 f5 41. Qc1 Qd5+ 42. Kh3 Kf7 43. Qc7+ Kg6
44. Qc8 Qe4 45. Qa6+ Kh5 46. Qa7 Qg4+ 47. Kg2 Qe2+ 48. Kg1 Qe1+ 49. Kg2
Qd2+ 50. Kh3 Qd3 51. Qf7+ Kh6 52. Qe7 Qe4 53. Qd6+ Kh5 54. Qd7 Qe5 55. Qf7+
Kh6 56. Qc4 g5 57. Qf7 f4 58. Qf8+ Kg6 59. Qg8+ Kf6 60. Qh8+ Ke6 61. Qg8+
Kd6 62. Qd8+ Kc6 63. Qc8+ Kb6 64. Qd8+ Kb5 65. Kg4 fxg3 66. Qd7+ Kb6 67.
Qd8+ Kc6 68. Qc8+ Kd5 69. Qd7+ Kc5 70. Qa7+ Kc6 71. Qa6+ Kd7 72. hxg3 Ke7
73. Qb7+ Kf6 74. Qc6+ Kg7 75. Qd7+ Kg8 76. Qd8+ Kf7 77. Qd7+ Kg6 78. Qc6+
Qf6 79. Qe4+ Kg7 80. Qb7+ Qf7 81. Qxf7+ Kxf7 82. Kxg5 Ke6 83. Kh6 Kf5 84.
Kxh7 Kg4 85. Kg6 Kxg3 86. Kf5 Kf3 87. Ke5 Ke3 88. Kd5 Kd2 89. Kc4 Kc2 90.
Kb5 Kb2 91. Kxa5 Kc3 92. Ka4 Kd4 93. Kxb4 Ke5 94. Kc5 Ke6 95. Kc6 Ke5 96.
b4 Kd4 97. b5 Kc3 98. b6 Kb4 99. Kd5 Ka3 100. b7 Kb2 101. Kd4 Kc2 102. b8=Q
Kd2 103. Qb2+ Kd1 104. Kd3 Ke1 105. Qe2#
{Xboard adjudication: Checkmate} 1-0

[swami]

Brilliant move, Miguel, keep it up! Gaviota seemed to have mastered the art of positional gain rather than materialistic one. In this particular case, It values initiative, space, positional superiority, mobility over the material. This is the kind of style that fascinates many comp chess hobbyists.

Perhaps its something to do with opening book since it happened in 2 games?

What did Rybka say to Nxd5? Does it give a normal or positive scoring?

Btw, has the Arena bug regarding check where older version of Gaviota send illegal move been fixed with the latest version of Arena 2.0 (or Gaviota?)

[michiguel]

Brilliant move, Miguel, keep it up! Gaviota seemed to have mastered the art of positional gain rather than materialistic one. In this particular case, It values initiative, space, positional superiority, mobility over the material. This is the kind of style that fascinates many comp chess hobbyists.

I am sure that Nd5 is wrong, but it is fun to watch.

Perhaps its something to do with opening book since it happened in 2 games?

No, I checked, Nd5 is the first move from the engine. It happens twice because it is one of the Noomen positions. I am playing matches with all the Nunn, Silver, and Noomen positions. Gaviota chooses Nd5, but in a deeper search prefers something else.

What did Rybka say to Nxd5? Does it give a normal or positive scoring?

[]
Rybka will think it is an illegal move
[]

Btw, has the Arena bug regarding check where older version of Gaviota send illegal move been fixed with the latest version of Arena 2.0 (or Gaviota?)

I don't know, I do not use Arena...

I just watched another interesting gambit played by Gaviota

[D]rnbr2k1/pp2p1bp/6p1/3P1p2/4P3/5N2/P2BBPPP/1R2K2R w K - 0 14

Gaviota just played f5!?. It would be great if black could retake with the Bishop but b7 is hanging. Phalanx played exf5 and Gaviota played... b6!? Very interesting. The combo f5-b6 makes sure that if white wants to keep d5, end up with the King stuck in the center.
It is interesting to see the pressure that black mounted on the center, even with two pawns down. Black won the endgame with a great active King.

Miguel

[Event "Computer Chess Game"]
[Site "avefenix"]
[Date "2009.04.01"]
[Round "2"]
[White "phalanx"]
[Black "Gaviota 0.59-test"]
[Result "0-1"]
[TimeControl "40/20"]

1. d4 Nf6 2. c4 g6 3. Nc3 d5 4. cxd5 Nxd5 5. e4 Nxc3 6. bxc3 Bg7 7. Nf3 c5
8. Rb1 O-O 9. Be2 cxd4 10. cxd4 Qa5+ 11. Qd2 Qxd2+ 12. Bxd2 Rd8 13. d5 f5
14. exf5 b6 15. fxg6 hxg6 16. Bc4 Ba6 17. Bb3 Nd7 18. d6+ Kf8 19. dxe7+
Kxe7 20. Bc2 Nc5 21. Bxg6 Rd6 22. Nh4 Nd3+ 23. Bxd3 Bxd3 24. Rc1 Kd7 25.
Kd1 Re8 26. Re1 Rxe1+ 27. Kxe1 Bb2 28. Be3 Bxc1 29. Bxc1 Rc6 30. Bf4 Rc2
31. Nf3 Re2+ 32. Kd1 Ba6 33. Bd2 Rxf2 34. Nh4 Bc4 35. a3 Kc6 36. Bc1 Kd5
37. Ke1 Re2+ 38. Kd1 Ke4 39. Bd2 Kd3 40. Nf3 Bb3+ 41. Kc1 Ke4 42. Nh4 Rf2
43. Bb4 Rc2+ 44. Kb1 a6 45. Bd6 Rd2 46. Bc7 Ba2+ 47. Kc1 Kd3 48. Bd8 Rc2+
49. Kd1 Bb3 50. Ke1 b5 51. Ba5 Bd5 52. Kf1 Ke3 53. Bb6+ Kf4 54. Bd8 Be4 55.
Kg1 Ra2 56. Be7 Rc2 57. Bd8 Kg4 58. Kf1 Rc8 59. Be7 Re8 60. h3+ Kg3 61.
Nf5+ Bxf5 62. Bd6+ Kh4 63. Kf2 Be4 64. Bg3+ Kh5 65. Bd6 Re6 66. Bf4 Rc6 67.
g3 Rc3 68. Bd6 Rd3 69. g4+ Kg5 70. Be7+ Kf4 71. g5 Rxh3 72. Ke1 Bf5 73. Kd1
a5 74. Kd2 Rd3+ 75. Kc1 Re3 76. Bd6+ Kxg5 77. Kb2 Be6 78. Bc5 Rb3+ 79. Kc2
Kf5 80. Bd6 Ke4 81. Bc5 Bc4 82. Kd2 Kd5 83. Be7 Kc6 84. Kc2 Re3 85. Bf8 Rb3
86. Be7 Kd5 87. Kd2 Rd3+ 88. Kc2 Rg3 89. Kb2 Rb3+ 90. Kc2 Re3 91. Bf8 Rf3
92. Be7 Rg3 93. Kb2 Rb3+ 94. Kc2 Re3 95. Bf8 Rd3 96. Be7 Kc6 97. Bf8 Re3
98. Kd2 Rf3 99. Be7 Rd3+ 100. Kc2 Rg3 101. Kb2 Re3 102. Bf8 Rf3 103. Be7
Re3 104. Bf8 Rb3+ 105. Kc2 Rf3 106. Be7 Re3 107. Bf8 Rf3 108. Be7 Rb3 109.
Bf8 Rh3 110. Kb2 Rh2+ 111. Kc3 Kd5 112. Be7 b4+ 113. axb4 axb4+ 114. Kxb4
Rh3 115. Ka4 Bb3+ 116. Kb5 Rh7 117. Bf6 Rb7+ 118. Ka6 Rb4 119. Ka5 Rf4 120.
Bg5 Rg4 121. Bd2 Kc6 122. Ka6 Ra4+ 123. Ba5 Bc4+ 124. Ka7 Rxa5+ 125. Kb8
Bd5 126. Kc8 Ra8#
{Xboard adjudication: Checkmate} 0-1

[AdminX]

Here is Rybka's assessment after 6 minutes thinking time of the position after "Nd5".

Hardware:
Intel Core 2 Quad Q6700 @ 2.66 Ghz
Windows Vista Ultimate 64-bit
8 GB Ram, 2048 Hash
3, 4, 5, and 6 Men Tablebases

[d]r4rk1/1bqnbppp/pp1ppn2/3N4/2PNP3/4B1P1/PP3PBP/2RQR1K1 b - - 0 1

Analysis by Deep Rybka 3 64-bit:

1...exd5 2.cxd5[] Qb8 3.Nc6 Bxc6 4.dxc6 Nc5 5.c7 Qa7 6.b4 Ncd7 7.Re2 Rac8 8.Rec2[] a5 9.a3 h5 10.Bd4 axb4 11.axb4 Qa4 12.Bh3 Qxb4 13.Bxf6[] Nxf6 14.Bxc8 Rxc8 15.e5 dxe5 16.Rd2 Kh7
=/+ (-0.62) Depth: 6 00:00:00 2kN
= (0.00) Depth: 21 00:06:57 83851kN


The move is clearly not a blunder but turns out to make for a interesting game. Looks like Gaviota has made a nice fine. Below is the final position reached at the end of Rybka's analysis:

[d]2r5/2P1bppk/1p3n2/4p2p/1q6/6P1/3R1P1P/2RQ2K1 w - - 0 17

[AdminX]

Here is Rybka's analysis after a hour of thinking time ,,,

Intel Core 2 Quad Q6700 @ 2.66 Ghz
Windows Vista Ultimate 64-bit
8 GB Ram, 2048 Hash
3, 4, 5, and 6 Men Tablebases

[d]r4rk1/1bqnbppp/pp1ppn2/3N4/2PNP3/4B1P1/PP3PBP/2RQR1K1 b - - 0 13

Analysis by Deep Rybka 3 64-bit:

13...exd5 14.cxd5[] Qb8 15.Nc6 Bxc6 16.dxc6 Nc5 17.c7 Qa7 18.Bf4 Rac8 19.e5[] dxe5[] 20.Bxe5 a5 21.a3 Ne6 22.Bh3 Kh8 23.Rc6 Qb7 24.Bg2 Qa6 25.Bf1 Qa8 26.Bg2
= (0.00) Depth: 15 00:00:00 7kN
= (0.00) Depth: 24 01:11:18 957mN

Position at the end of analysis:

[d]q1r2r1k/2P1bppp/1pR1nn2/p3B3/8/P5P1/1P3PBP/3QR1K1 b - - 0 26

[michiguel]

Here is Rybka's analysis after a hour of thinking time ,,,

Intel Core 2 Quad Q6700 @ 2.66 Ghz
Windows Vista Ultimate 64-bit
8 GB Ram, 2048 Hash
3, 4, 5, and 6 Men Tablebases

[d]r4rk1/1bqnbppp/pp1ppn2/3N4/2PNP3/4B1P1/PP3PBP/2RQR1K1 b - - 0 13

Analysis by Deep Rybka 3 64-bit:

13...exd5 14.cxd5[] Qb8 15.Nc6 Bxc6 16.dxc6 Nc5 17.c7 Qa7 18.Bf4 Rac8 19.e5[] dxe5[] 20.Bxe5 a5 21.a3 Ne6 22.Bh3 Kh8 23.Rc6 Qb7 24.Bg2 Qa6 25.Bf1 Qa8 26.Bg2
= (0.00) Depth: 15 00:00:00 7kN
= (0.00) Depth: 24 01:11:18 957mN

Position at the end of analysis:

[d]q1r2r1k/2P1bppp/1pR1nn2/p3B3/8/P5P1/1P3PBP/3QR1K1 b - - 0 26

Freaking unbelievable.

It would interesting to see a Montecarlo analysis. Computers were supposed to be extremely materialistic, but not anymore!

Miguel

[Dr.Wael Deeb]

It looks like your current version of Gaviota has something similar to the Deep Junior 7 playing style....
Eagrly waiting to your next release
Dr.D

[AdminX]

Freaking unbelievable.

It would interesting to see a Montecarlo analysis. Computers were supposed to be extremely materialistic, but not anymore!

Miguel

I took a small sampling of 400 games via Monte Carlo and here are the results.

Search Depth: 10

Number of Engines: 4

[d]r4rk1/1bqnbppp/pp1ppn2/3N4/2PNP3/4B1P1/PP3PBP/2RQR1K1 b - - 0 13

[Eelco de Groot]

For certain Blueberry would not play the move but in the end has to admit that it seems no less correct than other moves. Very good!

After 12. O-O

r4rk1/1bqnbppp/pp1ppn2/8/2PNP3/2N1B1P1/PP3PBP/2RQR1K1 w - -

Engine: Blueberry Beta 4 DM70 Build 421 (256 MB)
by F. Letouzey, T. Gaksch, E. de Groot

4 8:59 +0.16 13.f3 h5 14.Bh3 Rad8 15.Bxe6 fxe6
16.Nxe6 Qb8 17.Na4 Bc6 18.b4 Kh8
19.Nxd8 Bxd8 (138.005.532) 241

4 11:12 -0.20 13.Qd2 Rfd8 14.Bg5 Ne5 15.Na4 Nxc4
16.Nxb6 Qxb6 17.Rxc4 (162.375.922) 241
____________________________________________________________

4 8:59 +0.16 13.f3 h5 14.Bh3 Rad8 15.Bxe6 fxe6
16.Nxe6 Qb8 17.Na4 Bc6 18.b4 Kh8
19.Nxd8 Bxd8 (138.005.532) 238

4 14:02 -0.03 13.b3 Rad8 14.Qe2 Nc5 15.Bd2 e5
16.Nc2 Qd7 17.Nd5 Qe6 18.Kh1 Nxd5 (201.416.003) 238
____________________________________________________________

5 14:56 -0.01 13.f3 Ne5 14.Na4 Rfc8 15.Bh3 b5
16.cxb5 Qa5 (219.133.524) 246

5 20:10 -0.14 13.b3 Rad8 14.Qe2 Kh8 15.f3 Ne5
16.Bh3 Ng6 17.Na4 (298.822.252) 246

After 13. Nd5

[D]r4rk1/1bqnbppp/pp1ppn2/3N4/2PNP3/4B1P1/PP3PBP/2RQR1K1 b - -

Engine: Blueberry Beta 4 DM70 Build 421 (256 MB)
by F. Letouzey, T. Gaksch, E. de Groot

9 2:08 +0.81 13...exd5 14.cxd5 Qd8 15.Nc6 Qe8
16.Nxe7+ Qxe7 17.Rc7 Rfb8 18.Qb3 Ne8
19.Rcc1 Nc5 20.Qxb6 Nxe4 (36.442.284) 292

9 2:25 -1.20 13...Qd8 14.Nxf6+ Bxf6 15.e5 Bxg2
16.exf6 Bb7 17.fxg7 Re8 18.Bh6 Qf6
19.Qd2 Ne5 20.b4 Rad8 (42.650.893) 292
____________________________________________________________

10 4:32 +0.66 13...exd5 14.cxd5 Qd8 15.Nc6 Qe8
16.e5 Nxe5 17.Nxe7+ Qxe7 18.f4 Qd8
19.fxe5 dxe5 20.Qb3 Bxd5 21.Qxb6 Bxg2
22.Qxd8 Rfxd8 23.Kxg2 Rac8 24.Rxc8 Rxc8 (80.421.131) 248

10 24:43 -1.20 13...Qd8 14.Nxf6+ Bxf6 15.e5 Bxg2
16.exf6 Bb7 17.fxg7 Re8 18.Nf3 Rc8
19.Nh4 Qc7 20.Bd4 e5 21.Be3 Kxg7
22.Nf5+ Kh8 23.Qxd6 Qxd6 24.Nxd6 (369.034.806) 248
____________________________________________________________

11 27:24 +0.82 13...exd5 14.cxd5 Qd8 15.Nc6 Qe8
16.e5 Nxe5 17.Bxb6 Rc8 18.f4 Ng6
19.Nxe7+ Nxe7 20.Rxc8 Bxc8 21.Qe2 Nd7
22.Qxe7 Qxe7 23.Rxe7 Nxb6 (407.524.291) 251

11 29:46 -0.63 13...Qd8 14.Nxf6+ Nxf6 15.e5 dxe5
16.Bxb7 exd4 17.Qxd4 Bc5 18.Qc3 Rb8
19.Bxa6 Ng4 20.Bxc5 bxc5 21.h3 Ra8
22.Rcd1 (448.851.352) 251
____________________________________________________________

12 70:52 +0.11 13...exd5 14.cxd5 Qb8 15.Nc6 Bxc6
16.dxc6 Nc5 17.b4 Ne6 18.Bh3 Nc7
19.Qd3 d5 20.e5 Ne4 21.a3 a5 22.f3 axb4
23.axb4 Bxb4 24.fxe4 Bxe1 25.Rxe1 (931.440.068) 219

Up to and including move 16 this is similar to Rybka's line, Rybka plays 17. c7, but the eval only comes down with the last output so Blueberry sees only very late that Black is not so good after Gaviotas novelty, at least I think it might be something of a novelty? My HS for Rybka book still has a move, 13. f4, in this position after 12. .. O-O so we are not very far from the theory on the English Hedgehog. I have not looked into any databases if there are more games, if this 13. Nd5 maybe was played before. I don't think many, or even any, programs would play it !

Eelco

[AdminX]

For certain Blueberry would not play the move but in the end has to admit that it seems no less correct than other moves. Very good!

Up to and including move 16 this is similar to Rybka's line, Rybka plays 17. c7, but the eval only comes down with the last output so Blueberry sees only very late that Black is not so good after Gaviotas novelty, at least I think it might be something of a novelty? My HS for Rybka book still has a move, 13. f4, in this position after 12. .. O-O so we are not very far from the theory on the English Hedgehog. I have not looked into any databases if there are more games, if this 13. Nd5 maybe was played before. I don't think many, or even any, programs would play it !

Eelco

I located two games in my database, here they are ...

Note: In both games as is also the case in the Rybka analysis the theme seems to be all about that white c-pawn.

[Event "South Wales op 3rd"]
[Site "Caerleon"]
[Date "2006.07.10"]
[Round "4.9"]
[White "Cox, Tristan"]
[Black "Broomfield, Matthew"]
[Result "1/2-1/2"]
[ECO "A30"]
[WhiteElo "2062"]
[BlackElo "2325"]
[PlyCount "74"]
[EventDate "2006.07.08"]
[EventType "swiss"]
[EventRounds "9"]
[EventCountry "WLS"]
[Source "ChessBase"]
[SourceDate "2006.09.14"]

1. Nf3 b6 2. g3 Bb7 3. Bg2 c5 4. c4 Nf6 5. O-O e6 6. Nc3 Be7 7. Re1 a6 8. e4 d6
9. d4 cxd4 10. Nxd4 Qc7 11. Be3 Nbd7 12. Rc1 O-O 13. Nd5 exd5 14. cxd5 Qd8 15.
Nc6 Bxc6 16. dxc6 Nc5 17. b4 Ne6 18. c7 Qc8 19. e5 dxe5 20. Bxa8 Qxa8 21. Bxb6
e4 22. Qc2 Rc8 23. Qa4 Qb7 24. Ba5 Nd4 25. Re3 h5 26. Rc4 Qd5 27. Rxd4 Qxd4 28.
Qc6 Ng4 29. Qxe4 Bf6 30. Qxd4 Bxd4 31. Rd3 Ne5 32. Rd1 Nc6 33. a4 Kf8 34. b5
axb5 35. Bb4+ Nxb4 36. Rxd4 Nc6 37. axb5 Rxc7 1/2-1/2

[Event "Graz op"]
[Site "Graz"]
[Date "1997.??.??"]
[Round "1"]
[White "Riemelmoser, Oliver"]
[Black "Spelec, Darko"]
[Result "1-0"]
[ECO "A30"]
[WhiteElo "2230"]
[PlyCount "81"]
[EventDate "1997.10.??"]
[EventType "swiss"]
[EventRounds "9"]
[EventCountry "AUT"]
[Source "ChessBase"]
[SourceDate "1999.11.16"]

1. Nf3 Nf6 2. c4 b6 3. g3 Bb7 4. Bg2 e6 5. O-O c5 6. Nc3 d6 7. Re1 Nbd7 8. e4
a6 9. d4 cxd4 10. Nxd4 Qc7 11. Be3 Be7 12. Rc1 O-O 13. Nd5 exd5 14. cxd5 Qd8
15. Nc6 Bxc6 16. dxc6 Nc5 17. b4 Ne6 18. Qb3 Ng4 19. c7 Qd7 20. Bxb6 Qb5 21.
Ba5 Rac8 22. Qd5 Nc5 23. Qd2 Nd3 24. a4 Qh5 25. Qxd3 Qxh2+ 26. Kf1 Qh5 27. Bf3
Bg5 28. Bxg4 Qxg4 29. Rc6 f5 30. exf5 Qh3+ 31. Kg1 Rxf5 32. Qf1 Qh6 33. Qxa6
Rff8 34. Rxd6 Qh3 35. Qc4+ Kh8 36. Qe6 Qh5 37. b5 Rce8 38. Qc4 Qf3 39. Qf1 Bf4
40. Rxe8 Bxd6 41. c8=Q 1-0