Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Rybka Discussion / GM Benjamin vs. Rybka - Game 4
- - By lkaufman (*****) Date 2008-01-05 04:05 Edited 2008-01-05 04:08
[Event "Blitz:90'+30""]
[Site "Potomac"]
[Date "2008.01.04"]
[Round "4"]
[White "Benjamin, Joel"]
[Black "Rybka 2.3.2m14"]
[Result "1/2-1/2"]
[PlyCount "278"]
[TimeControl "5400+30"]

{1024MB, Shredder11.ctg, RYBKA36} 1. e4 {0} c5 {0} 2. Nf3 {8} e6 {0} 3. c3 {31}
Nf6 {0} 4. Bd3 {39} Nc6 {0} 5. O-O {0} d5 {115} 6. e5 {87} Nd7 {0} 7. Bb5 {228}
Qb6 {180} 8. Qa4 {492} Be7 {209} 9. d3 {560} O-O {393} 10. Bxc6 {147} Qxc6 {0}
11. Qxc6 {374} bxc6 {0} 12. c4 {110} a5 {32} 13. a4 {721} Rb8 {142} 14. Nbd2 {
62} g5 {160} 15. h3 {52} h5 {132} 16. Re1 {111} Kh7 {23} 17. g4 {847} Ba6 {0}
18. Rb1 {365} Rh8 {135} 19. b3 {327} Kg6 {34} 20. Bb2 {165} Bd8 {0} 21. Bc3 {96
} Rb7 {37} 22. Bb2 {235} Kg7 {179} 23. Ba3 {36} Bc7 {92} 24. Kg2 {128} Rb6 {197
} 25. Kg1 {36} Rbb8 {336} 26. Nf1 {122} Rb7 {115} 27. N1d2 {68} Kg6 {86} 28.
Kg2 {57} Bd8 {150} 29. Bb2 {69} Kg7 {195} 30. Ba3 {163} Rb6 {23} 31. Rbc1 {97}
Rb8 {132} 32. Rb1 {66} Kg6 {88} 33. Bb2 {50} Bc7 {100} 34. Kg1 {78} Kg7 {133}
35. Ba3 {62} Rhe8 {112} 36. Bb2 {164} Bd8 {93} 37. Ba3 {45} Kg6 {89} 38. Kg2 {
45} h4 {76} 39. Kg1 {74} Bb6 {89} 40. Rbc1 {48} Rb7 {40} 41. Kg2 {81} Rd8 {87}
42. Kg1 {48} Rg8 {86} 43. Kg2 {58} Rh8 {27} 44. Kg1 {61} Rhb8 {84} 45. Kg2 {51}
Re8 {32} 46. Kg1 {28} Bc7 {83} 47. Rb1 {224} Rc8 {82} 48. Kg2 {26} Rd8 {54} 49.
Kg1 {71} Ra7 {80} 50. Kg2 {27} Rg8 {51} 51. Kg1 {20} Rb7 {78} 52. Kg2 {34} Rh8
{43} 53. Kg1 {10} Rbb8 {76} 54. Kg2 {11} Rhe8 {64} 55. Kg1 {122} Rb7 {74} 56.
Kg2 {49} Reb8 {24} 57. Kg1 {47} Ra8 {112} 58. Kg2 {12} Bb6 {59} 59. Rbc1 {42}
Rc8 {71} 60. Kg1 {44} Rbb8 {69} 61. Kg2 {158} Nf8 {6} 62. Rb1 {210} Kh6 {58}
63. Bc1 {104} Bd8 {68} 64. Nf1 {387} Ng6 {24} 65. Be3 {133} Be7 {263} 66. N1d2
{42} Rg8 {164} 67. Kh2 {81} d4 {49} 68. Bxg5+ {28} Bxg5 {0} 69. Nxg5 {7} Kxg5 {
8} 70. Ne4+ {19} Kh6 {21} 71. Nxc5 {34} Bc8 {0} 72. Re4 {39} Rd8 {20} 73. f4 {
84} Nf8 {33} 74. Rf1 {150} Nh7 {14} 75. Rf2 {61} Kg7 {68} 76. f5 {41} Ng5 {59}
77. Ref4 {18} Rb4 {41} 78. Rf1 {80} Kh6 {59} 79. R1f2 {44} Rb8 {13} 80. Rf1 {52
} Rb6 {6} 81. R1f2 {11} Rb4 {47} 82. Rf1 {7} Rb8 {101} 83. R1f2 {8} Kg7 {12}
84. Rf1 {90} Rb4 {57} 85. R1f2 {12} Rb6 {39} 86. Rf1 {7} Kg8 {56} 87. R1f2 {17}
Rb8 {47} 88. Rf1 {33} Re8 {55} 89. f6 {217} Rd8 {25} 90. R1f2 {171} Kh7 {55}
91. Rg2 {135} Kg6 {70} 92. Rg1 {20} Rb7 {32} 93. Rg2 {33} Rb4 {53} 94. Rg1 {120
} Rb6 {0} 95. Rg2 {14} Bb7 {53} 96. Rg1 {52} Kh7 {0} 97. Rg2 {18} Rb4 {53} 98.
Rg1 {20} Ba8 {33} 99. Rg2 {18} Kg6 {53} 100. Rg1 {24} Rbb8 {52} 101. Rg2 {34}
Kh6 {51} 102. Rg1 {10} Bb7 {51} 103. Rg2 {16} Bc8 {50} 104. Rg1 {32} Rb6 {18}
105. Rg2 {6} Rb4 {50} 106. Rg1 {7} Kg6 {43} 107. Rg2 {7} Rb6 {49} 108. Rg1 {7}
Bb7 {41} 109. Rg2 {7} Rb4 {49} 110. Rg1 {13} Kh6 {35} 111. Rg2 {8} Rb6 {33}
112. Rg1 {7} Ba8 {40} 113. Rg2 {8} Kg6 {48} 114. Rg1 {8} Rdb8 {47} 115. Rb1 {47
} Kh6 {0} 116. Rb2 {25} Rb4 {23} 117. Rb1 {21} Rd8 {48} 118. Rg1 {19} Rbb8 {26}
119. Rg2 {64} Kg6 {56} 120. Rg1 {13} Rb6 {47} 121. Rg2 {9} Rb7 {46} 122. Rg1 {
28} Rb4 {46} 123. Rg2 {33} Rc8 {177} 124. Rg1 {62} Rcb8 {42} 125. Rb1 {21} R4b6
{20} 126. Rb2 {0} Kh6 {42} 127. Rb1 {60} Rb4 {97} 128. Rb2 {18} Rd8 {40} 129.
Rg2 {20} Kh7 {39} 130. Rg1 {36} Rdb8 {152} 131. Rb1 {15} Kg6 {524} 132. Rb2 {13
} R4b7 {65} 133. Rb1 {19} Kh6 {45} 134. Rb2 {20} Kh7 {44} 135. Rb1 {18} Rb6 {22
} 136. Rb2 {14} Kg8 {23} 137. Rb1 {24} Nh7 {0} 138. Rb2 {200} Nf8 {20} 139. Rb1
{46} Rb4 {24} 1/2-1/2
Parent - - By JhorAVi (***) [ph] Date 2008-01-05 04:19
Now i understand why the live game was aborted at move 55.   It's Because the remaining moves are humiliating :)
Parent - - By brunjes (**) [us] Date 2008-01-05 04:36
It was indeed painful to watch this.

The question I have is: how can Vas and his team come up with a way to prevent this type of thing from happening over and over? it's one thing to avoid a specific opening (don't play that line in the Sicilian again), but a GM can, over time (not in this match necessarily, but in general), become adept at reaching these types of positions. I hope Vas and company take as many months as they need after this match to develop an antidote to this type of play before releasing Rybka 3.

Games where lots of pawns (all of them for a big chunk of this game) are left on the board for a long time seem ripe for this type of play by a skilled player (i.e. a GM -- it never works for a patzer like me of course).

Roy
Parent - By turbojuice1122 (Gold) [us] Date 2008-01-05 05:10
But this is important, too--and thus the more important to keep developing stronger and stronger engines on stronger and stronger hardware.  We need to see if we can answer the question of whether it is always possible to do this against an objectively far stronger opponent.  Is it possible that a human not too much stronger than Kasparov and playing in the style of Kramnik could get a draw with white against God with this method?  This was basically an advanced version of the Father method.
Parent - By Vasik Rajlich (Silver) [hu] Date 2008-01-05 14:27
We have worked on this, and we are tinkering slightly during the match. Some early trends are apparent, but let's see how the rest of the games go. Then we'll have more to discuss.

Vas
Parent - - By Uri Blass (*****) [il] Date 2008-01-05 06:46
Maybe Joel could even win this game by 138.g5 but of course winning and drawing is the same for him so it was not his purpose.

Analysis by rybka after 138.g5

Benjamin,J - Rybka 2.3.2m14, Blitz:90'+30 2008
br4k1/5p1n/1rp1pP2/p1N1P1P1/P1Pp1R1p/1P1P3P/7K/1R6 b - - 0 1


Analysis by Rybka 2.3.2a 32-bit :

138...Rb8-d8 139.Rf4xh4
  =  (0.23)   Depth: 5   00:00:00
138...Rb8-d8 139.Rf4xh4 Nh7-f8
  =  (0.20)   Depth: 6   00:00:00  2kN
138...Rb8-d8 139.Rf4xh4 Nh7-f8 140.Kh2-g3
  ²  (0.32)   Depth: 7   00:00:00  3kN
138...Rb8-d8 139.g5-g6 Nh7-f8 140.g6xf7+ Kg8xf7 141.Rf4xh4 Kf7-e8
  +-  (2.31)   Depth: 8   00:00:00  12kN
138...Nh7-f8 139.Rf4xd4 Nf8-g6 140.Rb1-e1 Ba8-b7
  ²  (0.44)   Depth: 8   00:00:00  15kN
138...Nh7-f8 139.Rf4xd4 Nf8-g6 140.Rb1-e1 Ba8-b7 141.Rd4-d7
  ²  (0.44)   Depth: 9   00:00:00  17kN
138...Nh7-f8 139.Rf4xd4 Nf8-g6 140.Rb1-e1 Ba8-b7 141.Rd4-d7 Rb6-b4
  ²  (0.57)   Depth: 10   00:00:00  21kN
138...Nh7-f8 139.Rf4xd4 Nf8-g6 140.Rb1-e1 Ba8-b7 141.Rd4-d7 Rb6-b4 142.Re1-e4
  ²  (0.57)   Depth: 11   00:00:00  28kN
138...Nh7-f8 139.Rf4xd4 Nf8-g6 140.Rb1-e1 Ba8-b7 141.Rd4-d7 Rb6-b4 142.Re1-e4 Kg8-f8
  ²  (0.63)   Depth: 12   00:00:00  37kN
138...Nh7-f8 139.Rf4xd4 Nf8-g6 140.Rb1-e1 Ba8-b7 141.Rd4-d7 Rb6-b4 142.Re1-e4 Kg8-f8 143.Rd7-c7
  ²  (0.64)   Depth: 13   00:00:01  56kN
138...Nh7-f8 139.Rf4xd4 Nf8-g6 140.Rb1-e1 Ba8-b7 141.Rd4-d7 Rb6-b4 142.Re1-e4 Kg8-f8 143.Rd7-c7 Kf8-g8
  ²  (0.61)   Depth: 14   00:00:01  93kN
138...Nh7-f8 139.Rf4xd4 Nf8-g6 140.Rb1-e1 Ba8-b7 141.Rd4-d7 Rb6-b4 142.Kh2-g2 Ng6-f4+ 143.Kg2-f3 Nf4-g6 144.Kf3-g4 Kg8-f8
  ±  (0.81)   Depth: 15   00:00:02  176kN
138...Nh7-f8 139.Rf4xd4 Nf8-g6 140.Rb1-e1 Ba8-b7 141.Rd4-d7 Rb6-b4 142.Kh2-g2 Ng6-f4+ 143.Kg2-f3 Nf4-g6 144.Kf3-g4 Kg8-f8 145.Rd7-c7
  ±  (0.79)   Depth: 16   00:00:03  279kN
138...Nh7-f8 139.Rf4xd4 Nf8-g6 140.Rb1-e1 Ba8-b7 141.Rd4-d7 Kg8-f8 142.Rd7-c7 Kf8-e8 143.Re1-e4 Bb7-c8 144.Kh2-g2 Rb8-a8 145.Kg2-f3
  ±  (0.95)   Depth: 17   00:00:11  830kN
138...Nh7-f8 139.Rf4xd4 Nf8-g6 140.Rb1-e1 Ba8-b7 141.Rd4-d7 Kg8-f8 142.Rd7-c7 Kf8-e8 143.Re1-e4 Bb7-c8 144.Kh2-g2 Rb8-a8 145.Re4-d4
  ±  (1.06)   Depth: 18   00:00:20  1509kN
138...Nh7-f8 139.Rf4xd4 Nf8-g6 140.Rb1-e1 Ba8-b7 141.Rd4-d7 Kg8-f8 142.Rd7-c7 Kf8-e8 143.Re1-e4 Bb7-c8 144.Kh2-g2 Rb8-a8 145.Re4-d4
  ±  (1.06)   Depth: 19   00:00:32  2288kN
138...Nh7-f8 139.Rf4xd4 Nf8-g6 140.Rb1-e1 Ba8-b7 141.Rd4-d7 Kg8-f8 142.Rd7-c7 Kf8-e8 143.Re1-e4 Bb7-a6 144.Rc7-a7 Ba6-c8 145.Ra7xa5
  ±  (1.23)   Depth: 20   00:01:14  5366kN
138...Nh7-f8 139.Rf4xd4 Nf8-g6 140.Rb1-e1 Ba8-b7 141.Rd4-d7 Rb6-b4 142.Kh2-g2 Ng6-f4+ 143.Kg2-f3 Nf4-g6 144.Kf3-g4 Kg8-f8 145.Rd7-c7
  ±  (1.26)   Depth: 21   00:02:24  10163kN

(,  05.01.2008)
Parent - - By Mars (**) [de] Date 2008-01-05 11:05
I'm rather wondering if any move after 89.f6 was an insult to Joel. Maybe he is too nice to say that but probably it would be my feelings in his place.

Martin
Parent - By Permanent Brain (*****) Date 2008-01-05 11:27
I feel the same way, that is why I started this thread:

http://rybkaforum.net/cgi-bin/rybkaforum/topic_show.pl?tid=2955
Parent - - By Uri Blass (*****) [il] Date 2008-01-05 12:40
I think that there is no insult.
Maybe it is a draw after 89.f6(did not analyze it) but
It was not clear that the computer cannot win.
There was a possible idea of Nd7 and c5 that rybka did not try to play.

Maybe it does not work (did not analyze it and did not look at posts at this moment) but it is not something that the operators of rybka could be sure about in the game.

My feeling is the opposite to you and not playing after 89.f6 is an insult to people who watch the game.
Uri
Parent - - By JohnL (***) Date 2008-01-05 13:42
I agree, operators shouldn't interfere unless they are completely sure. And in that position it is not obvious, as Uri says Nd7 was still possible.

Also, if Joel decides to lock the game up, he has to blame himself if it gets to 50-move rule manouvering...
Parent - - By Mars (**) [de] Date 2008-01-05 15:12
No, it was clear that the knight couldn't move. Black can't afford giving up h4.

Re the insult to the spectators: I don't think they have a right to demand anything regarding the game. It's something between the opponents and has nothing to do with the money payed (payed not by the people watching, by the way).

Seeing Rybka moving pieces forth and back leaves a taste of disrespect for the human player.

Martin
Parent - By BB (****) [gb] Date 2008-01-05 15:35

> No, it was clear that the knight couldn't move. Black can't afford giving up h4.


What if Black first gets a rook to h8? [The d8 rook can move without d4 falling if there is an f3 fork]. For instance, from move 92, Black plays Bb7 then Ba8, then Rh8, Rbd8, Nh7, planning Nf8 and Nd7. Of course, White can do more to stop this, and it might not win in any case [White will play g5 and double on the fourth rank], but it could give some winning chances, and some care is needed from White.

> Seeing Rybka moving pieces forth and back leaves a taste of disrespect for the human player.


I think the hope was that Rybka would do something more than just shuffle around.
Parent - - By turbojuice1122 (Gold) [us] Date 2008-01-05 15:50
Any Nd7 idea loses because of g5!, after which white crashes through.  You saw an example of this above.  Game 4 was basically drawn in a lost position for black.  Any way that black tries moving that knight either drops the h4-pawn via a g5! sacrifice, after which the king falls, or drops the d4-pawn, after which white's pawns overwhelm in the center.

As you mention, since a win was the same as a draw for Joel, he had no need to play g5!, but it is shocking to me that Rybka 3 - epsilon would allow this.
Parent - - By Uri Blass (*****) [il] Date 2008-01-05 16:08
Rybka is playing for a win so it needs to do almost everything not to allow draw by the 50 move rule.
You cannot learn from the match that rybka 3-epsilon with contempt 0 allow g5.

Uri
Parent - - By turbojuice1122 (Gold) [us] Date 2008-01-05 16:11
But having a contempt of 100, maybe 200 on Larry's fast computer, would seem necessary to allow this.  I recall analyzing with other programs, and allowing this sort of break was often worth a 2-pawn advantage to white in many of the variations.  I think that any contempt beyond about 50 seems extremely dangerous...after all, the human might not see the obscure third repetition draw that may exist in a critical line.

Of course, an argument could be made that a draw being the same as a loss requires maximum contempt, but I think this would produce some extremely undesirable behavior, especially early in the middlegame.
Parent - By Uri Blass (*****) [il] Date 2008-01-05 16:15
The contempt does not have to be a constant and it is logical to increase the contempt when
the 50 move rule get closer.

The opponent may miss repetition but I do not expect the opponent to miss 50 move rule draw.

Uri
Parent - By Hetman (*****) Date 2008-01-05 16:27
Hi,

I do have similar opinion but I am not sponsor.

It is interesting what will be the public comments of the Rybka play in that position. The majority will be deciding if that was showing that someone is not understanding the position or is fighting to the end an is trying to use all chances. Human might be overworked.

It looks that battle is not only on the board but around it, too.  ;-). Like in the WCC matches :-). To the anihilation of the opponnent.

The revange of the human could be the remaining games played the same way. No risks, no offering or getting sacrifaces for the initiative and no  sharp play. The dull games. Trading pieces does not mean that GM does not play, GM is steering to the position where the plan and knowledge are dominating.

Rgds
Hetman
Parent - By lkaufman (*****) Date 2008-01-05 20:26
When we were about 8 moves away from the fifty move rule end of the game, Rybka started showing lines that indicated she might play "unsoundly" (due to the high contempt) to avoid the draw. But finally she decided these lines were just too awful. Anyway it proved to me that we were right to play until the end here.
Parent - By lkaufman (*****) Date 2008-01-05 20:15
Both Joel and Rybka thought White would be better after 138 g5, but it is rather moot because Rybka only allowed it due to high contempt setting, desperately trying to avoid fifty move rule draw. If Joel does win a game it may be only because of this.
Parent - - By Debaser (***) Date 2008-01-05 11:04
Could have been 77. ...f4 a way to open the game?

CT2007.1 anti-human likes this, but to be honest without anti-human too ;)
Parent - - By BB (****) [gb] Date 2008-01-05 11:38

> Could have been 77. ...f4 a way to open the game?


I think you mean 77... f6, but I agree that it keeps more air in the ball. I think a human White should still be able to draw, but it becomes more tricky.
Parent - By Debaser (***) Date 2008-01-05 14:58
Yes, sorry I was seeing the board from the black side ;)

And yes too, it is not a winning move, it is just a way to open the game and then wait the human mistake. If GM Benjamin makes no mistakes, he will win the match for sure.
Up Topic Rybka Support & Discussion / Rybka Discussion / GM Benjamin vs. Rybka - Game 4

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill