HannibalArasan, Leiden 2014
Moderators: hgm, Harvey Williamson, bob
Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.

 Posts: 3719
 Joined: Fri Mar 10, 2006 4:23 am
 Location: http://www.arasanchess.org
HannibalArasan, Leiden 2014
This was the most interesting game of the tournament for me  I didn't get to view it live but here is some afterthefact analysis:
[pgn]
[Event "CSVN Programmers Tournament 47"]
[Site "Leiden"]
[Date "2014.11.16"]
[Round "5"]
[White "Hannibal"]
[Black "Arasan"]
[Result "10"]
[ECO "C11"]
[PlyCount "175"]
[EventDate "2014.11.16"]
[WhiteClock "0:08:30"]
[BlackClock "0:06:06"]
1. e4 e6 2. d4 d5 3. Nc3 Nf6 4. e5 Nfd7 5. f4 c5 6. Nf3 Nc6 7. Be3 Be7 8. dxc5
Bxc5 9. Qd2 OO ({The most common.} 9... Qb6 {is preferred in Hiarc's book.})
10. OOO Qa5 11. Kb1 {rather unusual} (11. a3 $142 Be7 12. Kb1 a6 $14 {1/21/
2 (49) deepthroat (2593)EtaoinShrdlu (2624) InfinityChess INT 2014}) 11... Bb4
12. a3 Bxc3 13. Qxc3 Qxc3 14. bxc3 b6 $146 ({Predecessor (5):} 14... Nb6 15.
Bb5 Bd7 16. c4 Nb8 17. Bxb6 Bxb5 18. Bxa7 {1/21/2 (18) Kritz,L (2593)Luther,
T (2586) Triesen 2007}) 15. Nd4 Ne7 16. Re1 $6 (16. g4) 16... Nc5 17. g4 Rd8
18. Rg1 Bd7 19. Bd3 Rac8 20. Ne2 Nc6 21. Rg2 Na5 22. Bd4 Ne4 23. Kb2 Rc7 24. h4
Nb7 25. Ng3 Nec5 26. Ne2 Rdc8 27. Rh2 Na5 28. Ka2 a6 29. Rg2 Bb5 30. Rd1 Bc4+
31. Ka1 Bb5 32. Ka2 Na4 33. Rb1 Nc6 34. Bg1 Na7 35. Rb4 Bxd3 36. cxd3 b5 37.
Rb1 Rxc3 $2 {Up till this point Black was better. Arasan will get an
attractivelooking attack from this move but in fact it is not best.} (37...
Nc6 $142 {and Black preserves an advantage.}) 38. Nxc3 Nxc3+ 39. Kb3 Nxb1 40.
Bxa7 Rc3+ $2 (40... Rc7 $11) 41. Kb2 Rxd3 42. Kxb1 Rxa3 43. f5 $14 {3 pawns
are not enough compensation for the Knight.} Ra4 44. h5 h6 45. Bc5 Rc4 46. Bd6
a5 47. Rb2 Rxg4 48. Rxb5 exf5 49. Rb8+ Kh7 50. Rf8 Re4 51. Rxf7 f4 52. Kc2 a4
53. Kd2 a3 54. Ra7 a2 55. Rxa2 g5 56. Ra7+ Kg8 57. Ra8+ Kg7 58. Bf8+ Kg8 59.
Bxh6+ Kh7 60. Bf8 f3 61. Ra7+ Kg8 62. Bc5 Rxe5 63. Ra8+ Kf7 64. Rf8+ Ke6 65.
Re8+ Kf5 66. Rxe5+ Kxe5 67. h6 f2 68. Ke2 f1=Q+ 69. Kxf1 Kf6 70. Be3 g4 71. Kf2
d4 72. Bxd4+ Kg6 73. Bg7 Kh7 74. Kg3 Kg8 75. Kxg4 Kf7 76. Kf5 Ke7 77. h7 Kd6
78. h8=Q Kc5 79. Qc8+ Kb5 80. Bc3 Kb6 81. Bb4 Kb5 82. Qe6 Kxb4 83. Ke4 Kc3 84.
Qc6+ Kd2 85. Qc7 Ke1 86. Qh2 Kd1 87. Kd3 Kc1 88. Qc2# 10
[/pgn]
I tried Arasan on this position:
[D] 2r3k1/n1r2ppp/p3p3/1p1pP3/n4PPP/P1PP4/K3N1R1/1R4B1 b  
and it found Nc6 (which I think is best), but it took a long time on my quad (15 minutes). Arasan played on a 12core box in the tournament but even so, this was outside its search horizon.
Jon
[pgn]
[Event "CSVN Programmers Tournament 47"]
[Site "Leiden"]
[Date "2014.11.16"]
[Round "5"]
[White "Hannibal"]
[Black "Arasan"]
[Result "10"]
[ECO "C11"]
[PlyCount "175"]
[EventDate "2014.11.16"]
[WhiteClock "0:08:30"]
[BlackClock "0:06:06"]
1. e4 e6 2. d4 d5 3. Nc3 Nf6 4. e5 Nfd7 5. f4 c5 6. Nf3 Nc6 7. Be3 Be7 8. dxc5
Bxc5 9. Qd2 OO ({The most common.} 9... Qb6 {is preferred in Hiarc's book.})
10. OOO Qa5 11. Kb1 {rather unusual} (11. a3 $142 Be7 12. Kb1 a6 $14 {1/21/
2 (49) deepthroat (2593)EtaoinShrdlu (2624) InfinityChess INT 2014}) 11... Bb4
12. a3 Bxc3 13. Qxc3 Qxc3 14. bxc3 b6 $146 ({Predecessor (5):} 14... Nb6 15.
Bb5 Bd7 16. c4 Nb8 17. Bxb6 Bxb5 18. Bxa7 {1/21/2 (18) Kritz,L (2593)Luther,
T (2586) Triesen 2007}) 15. Nd4 Ne7 16. Re1 $6 (16. g4) 16... Nc5 17. g4 Rd8
18. Rg1 Bd7 19. Bd3 Rac8 20. Ne2 Nc6 21. Rg2 Na5 22. Bd4 Ne4 23. Kb2 Rc7 24. h4
Nb7 25. Ng3 Nec5 26. Ne2 Rdc8 27. Rh2 Na5 28. Ka2 a6 29. Rg2 Bb5 30. Rd1 Bc4+
31. Ka1 Bb5 32. Ka2 Na4 33. Rb1 Nc6 34. Bg1 Na7 35. Rb4 Bxd3 36. cxd3 b5 37.
Rb1 Rxc3 $2 {Up till this point Black was better. Arasan will get an
attractivelooking attack from this move but in fact it is not best.} (37...
Nc6 $142 {and Black preserves an advantage.}) 38. Nxc3 Nxc3+ 39. Kb3 Nxb1 40.
Bxa7 Rc3+ $2 (40... Rc7 $11) 41. Kb2 Rxd3 42. Kxb1 Rxa3 43. f5 $14 {3 pawns
are not enough compensation for the Knight.} Ra4 44. h5 h6 45. Bc5 Rc4 46. Bd6
a5 47. Rb2 Rxg4 48. Rxb5 exf5 49. Rb8+ Kh7 50. Rf8 Re4 51. Rxf7 f4 52. Kc2 a4
53. Kd2 a3 54. Ra7 a2 55. Rxa2 g5 56. Ra7+ Kg8 57. Ra8+ Kg7 58. Bf8+ Kg8 59.
Bxh6+ Kh7 60. Bf8 f3 61. Ra7+ Kg8 62. Bc5 Rxe5 63. Ra8+ Kf7 64. Rf8+ Ke6 65.
Re8+ Kf5 66. Rxe5+ Kxe5 67. h6 f2 68. Ke2 f1=Q+ 69. Kxf1 Kf6 70. Be3 g4 71. Kf2
d4 72. Bxd4+ Kg6 73. Bg7 Kh7 74. Kg3 Kg8 75. Kxg4 Kf7 76. Kf5 Ke7 77. h7 Kd6
78. h8=Q Kc5 79. Qc8+ Kb5 80. Bc3 Kb6 81. Bb4 Kb5 82. Qe6 Kxb4 83. Ke4 Kc3 84.
Qc6+ Kd2 85. Qc7 Ke1 86. Qh2 Kd1 87. Kd3 Kc1 88. Qc2# 10
[/pgn]
I tried Arasan on this position:
[D] 2r3k1/n1r2ppp/p3p3/1p1pP3/n4PPP/P1PP4/K3N1R1/1R4B1 b  
and it found Nc6 (which I think is best), but it took a long time on my quad (15 minutes). Arasan played on a 12core box in the tournament but even so, this was outside its search horizon.
Jon
Re: HannibalArasan, Leiden 2014
This is apparently something based on king safety. Crafty chooses Nc6 quickly and sticks with it. On my quadi7 iMac it changed to Nc3 after 1.6 seconds and stuck with it through 5 minutes. Eval hovering around 0.50 (favors black).jdart wrote:This was the most interesting game of the tournament for me  I didn't get to view it live but here is some afterthefact analysis:
[pgn]
[Event "CSVN Programmers Tournament 47"]
[Site "Leiden"]
[Date "2014.11.16"]
[Round "5"]
[White "Hannibal"]
[Black "Arasan"]
[Result "10"]
[ECO "C11"]
[PlyCount "175"]
[EventDate "2014.11.16"]
[WhiteClock "0:08:30"]
[BlackClock "0:06:06"]
1. e4 e6 2. d4 d5 3. Nc3 Nf6 4. e5 Nfd7 5. f4 c5 6. Nf3 Nc6 7. Be3 Be7 8. dxc5
Bxc5 9. Qd2 OO ({The most common.} 9... Qb6 {is preferred in Hiarc's book.})
10. OOO Qa5 11. Kb1 {rather unusual} (11. a3 $142 Be7 12. Kb1 a6 $14 {1/21/
2 (49) deepthroat (2593)EtaoinShrdlu (2624) InfinityChess INT 2014}) 11... Bb4
12. a3 Bxc3 13. Qxc3 Qxc3 14. bxc3 b6 $146 ({Predecessor (5):} 14... Nb6 15.
Bb5 Bd7 16. c4 Nb8 17. Bxb6 Bxb5 18. Bxa7 {1/21/2 (18) Kritz,L (2593)Luther,
T (2586) Triesen 2007}) 15. Nd4 Ne7 16. Re1 $6 (16. g4) 16... Nc5 17. g4 Rd8
18. Rg1 Bd7 19. Bd3 Rac8 20. Ne2 Nc6 21. Rg2 Na5 22. Bd4 Ne4 23. Kb2 Rc7 24. h4
Nb7 25. Ng3 Nec5 26. Ne2 Rdc8 27. Rh2 Na5 28. Ka2 a6 29. Rg2 Bb5 30. Rd1 Bc4+
31. Ka1 Bb5 32. Ka2 Na4 33. Rb1 Nc6 34. Bg1 Na7 35. Rb4 Bxd3 36. cxd3 b5 37.
Rb1 Rxc3 $2 {Up till this point Black was better. Arasan will get an
attractivelooking attack from this move but in fact it is not best.} (37...
Nc6 $142 {and Black preserves an advantage.}) 38. Nxc3 Nxc3+ 39. Kb3 Nxb1 40.
Bxa7 Rc3+ $2 (40... Rc7 $11) 41. Kb2 Rxd3 42. Kxb1 Rxa3 43. f5 $14 {3 pawns
are not enough compensation for the Knight.} Ra4 44. h5 h6 45. Bc5 Rc4 46. Bd6
a5 47. Rb2 Rxg4 48. Rxb5 exf5 49. Rb8+ Kh7 50. Rf8 Re4 51. Rxf7 f4 52. Kc2 a4
53. Kd2 a3 54. Ra7 a2 55. Rxa2 g5 56. Ra7+ Kg8 57. Ra8+ Kg7 58. Bf8+ Kg8 59.
Bxh6+ Kh7 60. Bf8 f3 61. Ra7+ Kg8 62. Bc5 Rxe5 63. Ra8+ Kf7 64. Rf8+ Ke6 65.
Re8+ Kf5 66. Rxe5+ Kxe5 67. h6 f2 68. Ke2 f1=Q+ 69. Kxf1 Kf6 70. Be3 g4 71. Kf2
d4 72. Bxd4+ Kg6 73. Bg7 Kh7 74. Kg3 Kg8 75. Kxg4 Kf7 76. Kf5 Ke7 77. h7 Kd6
78. h8=Q Kc5 79. Qc8+ Kb5 80. Bc3 Kb6 81. Bb4 Kb5 82. Qe6 Kxb4 83. Ke4 Kc3 84.
Qc6+ Kd2 85. Qc7 Ke1 86. Qh2 Kd1 87. Kd3 Kc1 88. Qc2# 10
[/pgn]
I tried Arasan on this position:
[D] 2r3k1/n1r2ppp/p3p3/1p1pP3/n4PPP/P1PP4/K3N1R1/1R4B1 b  
and it found Nc6 (which I think is best), but it took a long time on my quad (15 minutes). Arasan played on a 12core box in the tournament but even so, this was outside its search horizon.
Jon
It liked Rxc3 until it switched to Nc6, score difference was 0.10 (1/10th pawn) better for Nc6.

 Posts: 1762
 Joined: Tue Jun 02, 2009 10:14 pm
Re: HannibalArasan, Leiden 2014
does indeed to be something that could be correct with particular evaluation terms. komodo prefers Nc6 instantly & sticks with it 
Analysis by Komodo 8 64bit:
1...Nc6 2.Rc1
=/+ (0.44 ) Depth: 7 00:00:00 22kN
1...Rxc3 2.Nxc3
=/+ (0.68 ++) Depth: 7 00:00:00 28kN
=/+ (0.34 ) Depth: 9 00:00:00 72kN
1...Nc6 2.Rc1 h6 3.h5 Ne7 4.Bd4 Kf8 5.Rc2 Nc6 6.Be3 d4 7.cxd4 Kg8
=/+ (0.50) Depth: 9 00:00:00 90kN
= (0.30) Depth: 11 00:00:00 235kN
1...Rxc3 2.Nxc3
=/+ (0.36) Depth: 11 00:00:00 410kN
= (0.24 ) Depth: 12 00:00:00 566kN
1...Nc6 2.Rc1 h5 3.gxh5 Ne7 4.Bd4 Nf5 5.Rcg1 Kh7 6.Rg5 a5 7.Rxf5 exf5 8.Rb1 Kh6 9.Rg1 b4 10.axb4 axb4 11.Ng3 Nxc3+ 12.Kb3 Kh7 13.Nxf5 g6 14.Ng3 gxh5 15.Ra1 Nb5 16.Bb2 Rc2 17.Nxh5 Kh6 18.Nf6 Rd2 19.Ra6 Rxd3+ 20.Kxb4 Rc4+ 21.Ka5 Rb3 22.Rb6 Rxb2
=/+ (0.48 ++) Depth: 12 00:00:00 606kN
=/+ (0.67) Depth: 29 00:01:56 1568MN, tb=4005
quick mpv thru d21 
Analysis by Komodo 8 64bit:
1. =/+ (0.69): 1...Nc6 2.Rc1 h5 3.gxh5 Ne7 4.Bd4 Nf5 5.Rcg1 Kh7 6.Rg5 a5 7.R1g2 Rc6 8.Rxf5 exf5 9.Rg5 Nc5 10.Rxf5 Kg8 11.Bxc5 Rxc5 12.Rg5 Kh7 13.Kb3 d4 14.e6 fxe6 15.Rxc5 Rxc5 16.Nxd4 Rd5 17.Kc2 Rxh5 18.Nxe6 Rxh4 19.c4
2. =/+ (0.49): 1...g6 2.Rc1 Nc6 3.Kb3 Ne7 4.Be3 h6 5.Rgg1 Kg7 6.Rc2 Kh7 7.Rcc1 Nc6 8.Ka2 Kg8 9.Rg2 Ne7 10.Bd4 Kh7 11.Rgg1 Kg7 12.Rc2 Nc6 13.Be3
3. = (0.28): 1...h6 2.Rc1 Nc6 3.h5 Ne7 4.Bd4 Kf8 5.Rgg1 Nc6 6.Bf2 Kg8 7.Be3 Ne7 8.Bd4 Kh7 9.Rg2 Rf8 10.Bf2 Nc6 11.Be3 Rfc8 12.Rg3 Ne7 13.Bd4
Analysis by Komodo 8 64bit:
1...Nc6 2.Rc1
=/+ (0.44 ) Depth: 7 00:00:00 22kN
1...Rxc3 2.Nxc3
=/+ (0.68 ++) Depth: 7 00:00:00 28kN
=/+ (0.34 ) Depth: 9 00:00:00 72kN
1...Nc6 2.Rc1 h6 3.h5 Ne7 4.Bd4 Kf8 5.Rc2 Nc6 6.Be3 d4 7.cxd4 Kg8
=/+ (0.50) Depth: 9 00:00:00 90kN
= (0.30) Depth: 11 00:00:00 235kN
1...Rxc3 2.Nxc3
=/+ (0.36) Depth: 11 00:00:00 410kN
= (0.24 ) Depth: 12 00:00:00 566kN
1...Nc6 2.Rc1 h5 3.gxh5 Ne7 4.Bd4 Nf5 5.Rcg1 Kh7 6.Rg5 a5 7.Rxf5 exf5 8.Rb1 Kh6 9.Rg1 b4 10.axb4 axb4 11.Ng3 Nxc3+ 12.Kb3 Kh7 13.Nxf5 g6 14.Ng3 gxh5 15.Ra1 Nb5 16.Bb2 Rc2 17.Nxh5 Kh6 18.Nf6 Rd2 19.Ra6 Rxd3+ 20.Kxb4 Rc4+ 21.Ka5 Rb3 22.Rb6 Rxb2
=/+ (0.48 ++) Depth: 12 00:00:00 606kN
=/+ (0.67) Depth: 29 00:01:56 1568MN, tb=4005
quick mpv thru d21 
Analysis by Komodo 8 64bit:
1. =/+ (0.69): 1...Nc6 2.Rc1 h5 3.gxh5 Ne7 4.Bd4 Nf5 5.Rcg1 Kh7 6.Rg5 a5 7.R1g2 Rc6 8.Rxf5 exf5 9.Rg5 Nc5 10.Rxf5 Kg8 11.Bxc5 Rxc5 12.Rg5 Kh7 13.Kb3 d4 14.e6 fxe6 15.Rxc5 Rxc5 16.Nxd4 Rd5 17.Kc2 Rxh5 18.Nxe6 Rxh4 19.c4
2. =/+ (0.49): 1...g6 2.Rc1 Nc6 3.Kb3 Ne7 4.Be3 h6 5.Rgg1 Kg7 6.Rc2 Kh7 7.Rcc1 Nc6 8.Ka2 Kg8 9.Rg2 Ne7 10.Bd4 Kh7 11.Rgg1 Kg7 12.Rc2 Nc6 13.Be3
3. = (0.28): 1...h6 2.Rc1 Nc6 3.h5 Ne7 4.Bd4 Kf8 5.Rgg1 Nc6 6.Bf2 Kg8 7.Be3 Ne7 8.Bd4 Kh7 9.Rg2 Rf8 10.Bf2 Nc6 11.Be3 Rfc8 12.Rg3 Ne7 13.Bd4

 Posts: 3719
 Joined: Fri Mar 10, 2006 4:23 am
 Location: http://www.arasanchess.org
Re: HannibalArasan, Leiden 2014
Yes, I think Arasan likes the (apparent) king attack too much. Although generally it is not very attackhappy.
Jon
Jon
 Graham Banks
 Posts: 32365
 Joined: Sun Feb 26, 2006 9:52 am
 Location: Auckland, NZ
Re: HannibalArasan, Leiden 2014
What is apparent is that Arasan 17.4 is comfortably the strongest Arasan to date, so I hope that any changes that you make don't see a backwards step with the next release.jdart wrote:Yes, I think Arasan likes the (apparent) king attack too much. Although generally it is not very attackhappy.
Jon
My email addresses:
gbanksnz at gmail.com
gbanksnz at yahoo.co.nz
gbanksnz at gmail.com
gbanksnz at yahoo.co.nz
Re: HannibalArasan, Leiden 2014
Deuterium development version likes Rxc3 as in the game, because of the following position, it is about passers (connected and protected) and 3 pawns for a minor piece.jdart wrote:This was the most interesting game of the tournament for me  I didn't get to view it live but here is some afterthefact analysis:
[pgn]
[Event "CSVN Programmers Tournament 47"]
[Site "Leiden"]
[Date "2014.11.16"]
[Round "5"]
[White "Hannibal"]
[Black "Arasan"]
[Result "10"]
[ECO "C11"]
[PlyCount "175"]
[EventDate "2014.11.16"]
[WhiteClock "0:08:30"]
[BlackClock "0:06:06"]
1. e4 e6 2. d4 d5 3. Nc3 Nf6 4. e5 Nfd7 5. f4 c5 6. Nf3 Nc6 7. Be3 Be7 8. dxc5
Bxc5 9. Qd2 OO ({The most common.} 9... Qb6 {is preferred in Hiarc's book.})
10. OOO Qa5 11. Kb1 {rather unusual} (11. a3 $142 Be7 12. Kb1 a6 $14 {1/21/
2 (49) deepthroat (2593)EtaoinShrdlu (2624) InfinityChess INT 2014}) 11... Bb4
12. a3 Bxc3 13. Qxc3 Qxc3 14. bxc3 b6 $146 ({Predecessor (5):} 14... Nb6 15.
Bb5 Bd7 16. c4 Nb8 17. Bxb6 Bxb5 18. Bxa7 {1/21/2 (18) Kritz,L (2593)Luther,
T (2586) Triesen 2007}) 15. Nd4 Ne7 16. Re1 $6 (16. g4) 16... Nc5 17. g4 Rd8
18. Rg1 Bd7 19. Bd3 Rac8 20. Ne2 Nc6 21. Rg2 Na5 22. Bd4 Ne4 23. Kb2 Rc7 24. h4
Nb7 25. Ng3 Nec5 26. Ne2 Rdc8 27. Rh2 Na5 28. Ka2 a6 29. Rg2 Bb5 30. Rd1 Bc4+
31. Ka1 Bb5 32. Ka2 Na4 33. Rb1 Nc6 34. Bg1 Na7 35. Rb4 Bxd3 36. cxd3 b5 37.
Rb1 Rxc3 $2 {Up till this point Black was better. Arasan will get an
attractivelooking attack from this move but in fact it is not best.} (37...
Nc6 $142 {and Black preserves an advantage.}) 38. Nxc3 Nxc3+ 39. Kb3 Nxb1 40.
Bxa7 Rc3+ $2 (40... Rc7 $11) 41. Kb2 Rxd3 42. Kxb1 Rxa3 43. f5 $14 {3 pawns
are not enough compensation for the Knight.} Ra4 44. h5 h6 45. Bc5 Rc4 46. Bd6
a5 47. Rb2 Rxg4 48. Rxb5 exf5 49. Rb8+ Kh7 50. Rf8 Re4 51. Rxf7 f4 52. Kc2 a4
53. Kd2 a3 54. Ra7 a2 55. Rxa2 g5 56. Ra7+ Kg8 57. Ra8+ Kg7 58. Bf8+ Kg8 59.
Bxh6+ Kh7 60. Bf8 f3 61. Ra7+ Kg8 62. Bc5 Rxe5 63. Ra8+ Kf7 64. Rf8+ Ke6 65.
Re8+ Kf5 66. Rxe5+ Kxe5 67. h6 f2 68. Ke2 f1=Q+ 69. Kxf1 Kf6 70. Be3 g4 71. Kf2
d4 72. Bxd4+ Kg6 73. Bg7 Kh7 74. Kg3 Kg8 75. Kxg4 Kf7 76. Kf5 Ke7 77. h7 Kd6
78. h8=Q Kc5 79. Qc8+ Kb5 80. Bc3 Kb6 81. Bb4 Kb5 82. Qe6 Kxb4 83. Ke4 Kc3 84.
Qc6+ Kd2 85. Qc7 Ke1 86. Qh2 Kd1 87. Kd3 Kc1 88. Qc2# 10
[/pgn]
I tried Arasan on this position:
[D] 2r3k1/n1r2ppp/p3p3/1p1pP3/n4PPP/P1PP4/K3N1R1/1R4B1 b  
and it found Nc6 (which I think is best), but it took a long time on my quad (15 minutes). Arasan played on a 12core box in the tournament but even so, this was outside its search horizon.
Jon
[d]6k1/B4ppp/p3p3/1p1pP3/5PPP/r7/6R1/1K6 w   0 7
Somehow black should at least get a draw here by exchanging those white's 4 pawns. But this is the kind of position that really tests engine's efficient search and evaluation estimates. Those passers are very tempting. Things like do you apply pruning or perhaps aggressive reduction in this kind of position? I have this kind of idea but have not implemented yet  to extend the search when material imbalance exists on the board. Say exchange sac, 3p vs 1 minor, 1p+r vs 2 minors and others.

 Posts: 3719
 Joined: Fri Mar 10, 2006 4:23 am
 Location: http://www.arasanchess.org
Re: HannibalArasan, Leiden 2014
Passers are tricky. These are not that far advanced, and the opposing King is in front of them. Plus of course White has a whole extra piece. With those considerations in mind Black should not be better. And in fact in the game White was able to fairly easily play against the passers.
Jon
Jon
Re: HannibalArasan, Leiden 2014
Gaviota seems to like d4jdart wrote:This was the most interesting game of the tournament for me  I didn't get to view it live but here is some afterthefact analysis:
[pgn]
[Event "CSVN Programmers Tournament 47"]
[Site "Leiden"]
[Date "2014.11.16"]
[Round "5"]
[White "Hannibal"]
[Black "Arasan"]
[Result "10"]
[ECO "C11"]
[PlyCount "175"]
[EventDate "2014.11.16"]
[WhiteClock "0:08:30"]
[BlackClock "0:06:06"]
1. e4 e6 2. d4 d5 3. Nc3 Nf6 4. e5 Nfd7 5. f4 c5 6. Nf3 Nc6 7. Be3 Be7 8. dxc5
Bxc5 9. Qd2 OO ({The most common.} 9... Qb6 {is preferred in Hiarc's book.})
10. OOO Qa5 11. Kb1 {rather unusual} (11. a3 $142 Be7 12. Kb1 a6 $14 {1/21/
2 (49) deepthroat (2593)EtaoinShrdlu (2624) InfinityChess INT 2014}) 11... Bb4
12. a3 Bxc3 13. Qxc3 Qxc3 14. bxc3 b6 $146 ({Predecessor (5):} 14... Nb6 15.
Bb5 Bd7 16. c4 Nb8 17. Bxb6 Bxb5 18. Bxa7 {1/21/2 (18) Kritz,L (2593)Luther,
T (2586) Triesen 2007}) 15. Nd4 Ne7 16. Re1 $6 (16. g4) 16... Nc5 17. g4 Rd8
18. Rg1 Bd7 19. Bd3 Rac8 20. Ne2 Nc6 21. Rg2 Na5 22. Bd4 Ne4 23. Kb2 Rc7 24. h4
Nb7 25. Ng3 Nec5 26. Ne2 Rdc8 27. Rh2 Na5 28. Ka2 a6 29. Rg2 Bb5 30. Rd1 Bc4+
31. Ka1 Bb5 32. Ka2 Na4 33. Rb1 Nc6 34. Bg1 Na7 35. Rb4 Bxd3 36. cxd3 b5 37.
Rb1 Rxc3 $2 {Up till this point Black was better. Arasan will get an
attractivelooking attack from this move but in fact it is not best.} (37...
Nc6 $142 {and Black preserves an advantage.}) 38. Nxc3 Nxc3+ 39. Kb3 Nxb1 40.
Bxa7 Rc3+ $2 (40... Rc7 $11) 41. Kb2 Rxd3 42. Kxb1 Rxa3 43. f5 $14 {3 pawns
are not enough compensation for the Knight.} Ra4 44. h5 h6 45. Bc5 Rc4 46. Bd6
a5 47. Rb2 Rxg4 48. Rxb5 exf5 49. Rb8+ Kh7 50. Rf8 Re4 51. Rxf7 f4 52. Kc2 a4
53. Kd2 a3 54. Ra7 a2 55. Rxa2 g5 56. Ra7+ Kg8 57. Ra8+ Kg7 58. Bf8+ Kg8 59.
Bxh6+ Kh7 60. Bf8 f3 61. Ra7+ Kg8 62. Bc5 Rxe5 63. Ra8+ Kf7 64. Rf8+ Ke6 65.
Re8+ Kf5 66. Rxe5+ Kxe5 67. h6 f2 68. Ke2 f1=Q+ 69. Kxf1 Kf6 70. Be3 g4 71. Kf2
d4 72. Bxd4+ Kg6 73. Bg7 Kh7 74. Kg3 Kg8 75. Kxg4 Kf7 76. Kf5 Ke7 77. h7 Kd6
78. h8=Q Kc5 79. Qc8+ Kb5 80. Bc3 Kb6 81. Bb4 Kb5 82. Qe6 Kxb4 83. Ke4 Kc3 84.
Qc6+ Kd2 85. Qc7 Ke1 86. Qh2 Kd1 87. Kd3 Kc1 88. Qc2# 10
[/pgn]
I tried Arasan on this position:
[D] 2r3k1/n1r2ppp/p3p3/1p1pP3/n4PPP/P1PP4/K3N1R1/1R4B1 b  
and it found Nc6 (which I think is best), but it took a long time on my quad (15 minutes). Arasan played on a 12core box in the tournament but even so, this was outside its search horizon.
Jon
Code: Select all
161629513 25 212.0 +0.47 d4 2.Bxd4 Nc6 3.Be3 Ne7 4.Bd4 Nd5 5.Rb3
Naxc3+ 6.Nxc3 Nxc3+ 7.Bxc3 Rxc3 8.Rd2
Rc2+ 9.Rxc2 Rxc2+ 10.Rb2 Rxb2+ 11.Kxb2
h5 12.gxh5 Kh7 13.Kc3 Kh6 14.Kb4 Kxh5
15.Ka5 Kg4 16.Kxa6 Kxf4 17.Kxb5 Kxe5
Miguel
 Eelco de Groot
 Posts: 4058
 Joined: Sun Mar 12, 2006 1:40 am
 Location: Groningen
Re: HannibalArasan, Leiden 2014
A Rainbow Serpent at depth 46:
2r3k1/n1r2ppp/p3p3/1p1pP3/n4PPP/P1PP4/K3N1R1/1R4B1 b  
Engine: Sf20141112_005 MOD MP (512 MB)
by Tord Romstad, Marco Costalba and Joona Kiiski
21/27 0:02 +0.22 37...Nc6 38.Rc1 h5 39.gxh5 Ne7
40.Kb3 Nf5 41.Bf2 Kf8 42.Rgg1 Ke7
43.Rg5 d4 44.Rxf5 Nc5+ 45.Kc2 exf5
46.h6 gxh6 47.Nxd4 Ne4 48.Nxf5+ Ke6 (5.623.546) 2721
.
.
.
.
27/33 0:16 +0.34 37...Nc6 38.Rc1 d4 39.cxd4 Ne7
40.Rxc7 Rxc7 41.Kb3 Nd5 42.f5 a5
43.h5 Nac3 44.Kb2 Nd1+ 45.Kb1 Kf8
46.fxe6 fxe6 47.g5 Kf7 48.Rg3 N1c3+
49.Nxc3 Nxc3+ 50.Kb2 Nd1+ (48.401.876) 2969
28/34 0:20 +0.27 37...Nc6 38.Rc1 (59.969.573) 2988
28/34 0:21 +0.34++ 37...Nc6 (64.483.678) 3003
28/34 0:22 +0.43++ 37...Nc6 (66.839.311) 3003
28/34 0:23 +0.57++ 37...Nc6 (71.111.960) 3007
28/34 0:26 +0.46 37...Nc6 38.Rc1 (81.733.944) 3029
28/34 0:28 +0.61 37...Nc6 38.Rc1 h5 39.gxh5 Ne7
40.Bd4 Nf5 41.Rcg1 Kh8 42.Rg5 (86.154.599) 3031
.
.
.
.
36/47 8:14 +0.58 37...Nc6 38.Rc1 d4 39.cxd4 Ne7
40.Rxc7 Rxc7 41.Kb1 Nc3+ 42.Kb2 Ned5
43.f5 Nd1+ 44.Kb1 h6 45.fxe6 fxe6
46.g5 hxg5 47.hxg5 N1c3+ 48.Kb2 Nxe2
49.Rxe2 Kf7 50.Rg2 Kg6 (1.571.346.745) 3179
.
.
.
.
45/59 247:02 +0.60 37...Nc6 38.Rc1 d4 39.Bxd4 Nxd4
40.Nxd4 Nxc3+ 41.Kb3 Nd5 42.Rxc7 Rxc7
43.Ne2 Kf8 44.Rg3 g6 45.h5 Ke7
46.Rf3 a5 47.hxg6 hxg6 48.a4 bxa4+
49.Kxa4 Rc2 50.Ng3 Ra2+ (48.705.250.191) 3285
46/59 285:14 +0.53 37...Nc6 38.Rc1 (56.483.078.497) 3300
46/59 308:42 +0.47 37...Nc6 38.Rc1 (60.883.142.632) 3286
46/59 330:26 +0.38 37...Nc6 38.Rc1 (64.973.349.244) 3277
46/59 337:13 +0.46++ 37...Nc6 (66.355.378.899) 3279
46/59 416:16 +0.43 37...Nc6 38.Rc1 d4 39.c4 bxc4
40.Rxc4 Nb6 41.Rc5 Nd5 42.Nxd4 Nxd4
43.Rxc7 Rxc7 44.Bxd4 Nxf4 45.Rb2 h6
46.Rb8+ Kh7 47.Kb3 Nxd3 48.Ra8 Nc1+
49.Kb4 Rc6 50.a4 Ne2 (82.037.775.514) 3284
best move: Na7c6 time: 431:07.695 min n/s: 3.284.622 nodes: 85.019.091.474
It looks like 37... d4 and 37... Nc6 might more or less transpose, although d4 Bxd4 is maybe not so good anymore after first Nc6.
I also wanted to show that there is some search instability, Fail High, Fail Low, Fail High again. Probably due to the aspiration windows if you start moving both alpha and beta. But this is not standard Stockfish so I don't know if you don't see this in Stockfish and time to depth should not be adversely affected as long as the re searches are with very small windows, almost like a null window search then. So the small elo gain might still be there. I do return here on a TT entry found in a PV node so that means not 100% accurate PVs anymore.
2r3k1/n1r2ppp/p3p3/1p1pP3/n4PPP/P1PP4/K3N1R1/1R4B1 b  
Engine: Sf20141112_005 MOD MP (512 MB)
by Tord Romstad, Marco Costalba and Joona Kiiski
21/27 0:02 +0.22 37...Nc6 38.Rc1 h5 39.gxh5 Ne7
40.Kb3 Nf5 41.Bf2 Kf8 42.Rgg1 Ke7
43.Rg5 d4 44.Rxf5 Nc5+ 45.Kc2 exf5
46.h6 gxh6 47.Nxd4 Ne4 48.Nxf5+ Ke6 (5.623.546) 2721
.
.
.
.
27/33 0:16 +0.34 37...Nc6 38.Rc1 d4 39.cxd4 Ne7
40.Rxc7 Rxc7 41.Kb3 Nd5 42.f5 a5
43.h5 Nac3 44.Kb2 Nd1+ 45.Kb1 Kf8
46.fxe6 fxe6 47.g5 Kf7 48.Rg3 N1c3+
49.Nxc3 Nxc3+ 50.Kb2 Nd1+ (48.401.876) 2969
28/34 0:20 +0.27 37...Nc6 38.Rc1 (59.969.573) 2988
28/34 0:21 +0.34++ 37...Nc6 (64.483.678) 3003
28/34 0:22 +0.43++ 37...Nc6 (66.839.311) 3003
28/34 0:23 +0.57++ 37...Nc6 (71.111.960) 3007
28/34 0:26 +0.46 37...Nc6 38.Rc1 (81.733.944) 3029
28/34 0:28 +0.61 37...Nc6 38.Rc1 h5 39.gxh5 Ne7
40.Bd4 Nf5 41.Rcg1 Kh8 42.Rg5 (86.154.599) 3031
.
.
.
.
36/47 8:14 +0.58 37...Nc6 38.Rc1 d4 39.cxd4 Ne7
40.Rxc7 Rxc7 41.Kb1 Nc3+ 42.Kb2 Ned5
43.f5 Nd1+ 44.Kb1 h6 45.fxe6 fxe6
46.g5 hxg5 47.hxg5 N1c3+ 48.Kb2 Nxe2
49.Rxe2 Kf7 50.Rg2 Kg6 (1.571.346.745) 3179
.
.
.
.
45/59 247:02 +0.60 37...Nc6 38.Rc1 d4 39.Bxd4 Nxd4
40.Nxd4 Nxc3+ 41.Kb3 Nd5 42.Rxc7 Rxc7
43.Ne2 Kf8 44.Rg3 g6 45.h5 Ke7
46.Rf3 a5 47.hxg6 hxg6 48.a4 bxa4+
49.Kxa4 Rc2 50.Ng3 Ra2+ (48.705.250.191) 3285
46/59 285:14 +0.53 37...Nc6 38.Rc1 (56.483.078.497) 3300
46/59 308:42 +0.47 37...Nc6 38.Rc1 (60.883.142.632) 3286
46/59 330:26 +0.38 37...Nc6 38.Rc1 (64.973.349.244) 3277
46/59 337:13 +0.46++ 37...Nc6 (66.355.378.899) 3279
46/59 416:16 +0.43 37...Nc6 38.Rc1 d4 39.c4 bxc4
40.Rxc4 Nb6 41.Rc5 Nd5 42.Nxd4 Nxd4
43.Rxc7 Rxc7 44.Bxd4 Nxf4 45.Rb2 h6
46.Rb8+ Kh7 47.Kb3 Nxd3 48.Ra8 Nc1+
49.Kb4 Rc6 50.a4 Ne2 (82.037.775.514) 3284
best move: Na7c6 time: 431:07.695 min n/s: 3.284.622 nodes: 85.019.091.474
It looks like 37... d4 and 37... Nc6 might more or less transpose, although d4 Bxd4 is maybe not so good anymore after first Nc6.
I also wanted to show that there is some search instability, Fail High, Fail Low, Fail High again. Probably due to the aspiration windows if you start moving both alpha and beta. But this is not standard Stockfish so I don't know if you don't see this in Stockfish and time to depth should not be adversely affected as long as the re searches are with very small windows, almost like a null window search then. So the small elo gain might still be there. I do return here on a TT entry found in a PV node so that means not 100% accurate PVs anymore.
Debugging is twice as hard as writing the code in the first
place. Therefore, if you write the code as cleverly as possible, you
are, by definition, not smart enough to debug it.
 Brian W. Kernighan
place. Therefore, if you write the code as cleverly as possible, you
are, by definition, not smart enough to debug it.
 Brian W. Kernighan
Re: HannibalArasan, Leiden 2014
Latest SF chooses Nc6 (starting rather early in the search):jdart wrote: I tried Arasan on this position:
[D] 2r3k1/n1r2ppp/p3p3/1p1pP3/n4PPP/P1PP4/K3N1R1/1R4B1 b  
and it found Nc6 (which I think is best), but it took a long time on my quad (15 minutes). Arasan played on a 12core box in the tournament but even so, this was outside its search horizon.
Jon
Code: Select all
0.50 [+] [*] 1... Nc6 2. Rc1 d4 3. c4 bxc4 4. Rxc4 Nb6 5. Rc5 Nd5 6. Nxd4 Nxd4 7. Rxc7 Rxc7 8. Bxd4 Nxf4 9. Rb2 h5 10. gxh5 Rd7 11. Rb8+ Kh7 12. Bb2 Nxd3 13. Bc3 Rc7 14. Kb3 Nc1+ 15. Kb2 Rc4 16. Rb7 Nd3+ 17. Kb3 Rxh4 18. Rxf7 Nc5+ 19. Kc2 Rh2+ 20. Kd1 Rxh5 21. Re7 Ne4 22. Bd4 Ng5 23. Kc2 Kg6 24. Bc3 Rh2+ 25. Kd3 Rh3+ 26. Kc4 Rf3 27. Kd4 Rg3 28. Re8 Rh3 29. Ra8 Nf7 30. Kc4 Kf5 31. Rf8 Rh4+ 32. Kb3 Kg6 33. Ra8 (depth 45, 0:41:50)