Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Rybka Discussion / PGNs of game 3 and 4 with Rybka evaluation, Zappa won both
- - By Alkelele (****) Date 2007-09-23 01:48 Edited 2007-09-23 01:51
A few comments on the games:

In game 3, we left book in a reasonable position, but unfortunately a position of a type where Rybka immediately displayed her common misappreciation of white's d4-d5 push in this Ruy Lopez structure with exd4 and cxd4. 15...Nc6 (first move out of book) was a disaster, and the second move, 16...Na5 (instead of 16...Ne5) just gave Rybka a pretty hopeless position. Zappa's 22.f3 was nice, cementing black's horrible bishop on g6. We played this game with the newest and strongest version, including TBs. However, some reports led to doubts about whether there might be some problem with this version, so we decided to go back to the intermediate version (+TBs) for the 4th game. The truth most probably was that black just got a very bad position right away and that no engine version would have had decent chances to hold the position.

Game 4 was an epic tragedy. We got a practically winning position straight out of book, a pre-match assessment of ours that the course of the game quickly validated. Rybka was just well on top of Zappa in the early stages after book, seeing the huge advantage on her own much earlier. But, after having made steady progress, Rybka failed to realize the strength of 52.Qe1, which probably wins the game pretty much on the spot (while Zappa did see it before playing gxh5 and was giving scores like +2.7). Instead, we got a blocked position where Rybka couldn't make reasonable progress (after f3-f4-f5). Most people were willing to call it a draw and a day, but Erdo wisely insisted on playing on and was heavily rewarded for this. The first time Rybka reached the 50-move limit, she gave away a pawn as expected to avoid an instant draw, and further on, well, I really don't know, it was just ugly :-) It was pretty shocking that even when black had "just" achieved full equality to judge by the engine evaluations, that is, when white had queen+two pawns vs. two rooks+bishop, white was probably close to lost. An extremely interesting endgame, and a shocking end to an epic battle.(Btw, both engines played with 5 man TBs and some 6 man TBs, but not the relevant 6 man TB).

We have our work cut out for us in the coming days, for sure.

[Event "Clash of the Computer Titans"]
[Site "?"]
[Date "2007.09.22"]
[Round "3"]
[White "Zappa "]
[Black "Rybka"]
[Result "1-0"]
[ECO "C91"]
[Annotator "Rybka"]
[PlyCount "131"]
[TimeControl "3600+20"]

{1024MB, MexicoB.ctg} 1. e4 {0} e5 {B/0 0} 2. Nf3 {13} Nc6 {B/0 0} 3. Bb5 {11}
a6 {B/0 0} 4. Ba4 {10} Nf6 {B/0 0} 5. O-O {10} Be7 {B/0 0} 6. Re1 {9} b5 {B/0 0
} 7. Bb3 {11} d6 {B/0 0} 8. c3 {9} O-O {B/0 0} 9. d4 {10} Bg4 {B/0 0} 10. Be3 {
15} exd4 {B/0 0} 11. cxd4 {9} Na5 {B/0 0} 12. Bc2 {10} c5 {B/0 0} 13. h3 {8}
Bh5 {B/0 0} 14. g4 {9} Bg6 {Last book move. B/0 0} 15. Nbd2 {Last book move. 12
} Nc6 {0.00/15 86} 16. d5 {(Rc1) 193} Na5 {0.06/18 193} 17. a3 {166} Re8 {
0.08/20 690} 18. b4 {157} Nb7 {0.08/18 1} 19. a4 {(bxc5) 302} cxb4 {0.04/16 64}
20. Nd4 {(axb5) 114} Nc5 {0.00/15 62} 21. Nc6 {(axb5) 159} Qc8 {0.00/15 62} 22.
f3 {(g5) 111} Nfd7 {0.13/16 94} 23. axb5 {(Rb1) 131} Bh4 {0.27/17 166} 24. Rf1
{196} Bf6 {0.23/17 1} 25. Rb1 {208} axb5 {0.30/17 1} 26. Rxb4 {112} Nb8 {
0.40/20 954} 27. Nxb8 {12} Rxb8 {0.38/20 102} 28. Qb1 {9} Qd7 {0.48/19 75} 29.
Bd3 {145} Nxd3 {0.48/18 1} 30. Qxd3 {11} Rec8 {0.51/20 52} 31. Rfb1 {116} Rc3 {
0.51/18 1} 32. Rxb5 {132} Rbc8 {0.59/19 1} 33. Qe2 {28} h6 {0.59/18 40} 34. Rb8
{(R5b4) 207} Kh7 {0.72/17 55} 35. Rxc8 {150} Qxc8 {0.88/19 1} 36. f4 {114} Kh8
{0.88/19 1} 37. Kg2 {242} Qe8 {0.93/18 1} 38. f5 {(Rb4) 213} Bh7 {1.10/17 14}
39. Rb6 {224} Qe5 {0.96/18 1} 40. Bf2 {127} Rc2 {1.29/21 1} 41. Qd3 {167} Ra2 {
1.27/21 1} 42. Rb1 {(Nf3) 159} Qf4 {1.17/20 96} 43. Qe3 {56} Qe5 {1.27/20 21}
44. Nf3 {53} Qe8 {1.27/18 1} 45. Qb3 {51} Ra8 {1.27/17 1} 46. Bg3 {85} Qd8 {
1.26/18 86} 47. Rc1 {(Qb6) 8} Qe7 {1.31/17 49} 48. Re1 {(Rc4) 12} Qd8 {
1.33/21 329} 49. Rc1 {(Re2) 8} Qe7 {0.00/52 14} 50. Rc4 {(Bxd6) 154} Bg8 {
1.34/17 35} 51. h4 {(Bf4) 9} Qd8 {1.59/18 91} 52. Rc6 {9} Be7 {1.59/16 20} 53.
g5 {(Bf4) 107} hxg5 {1.85/17 87} 54. hxg5 {9} Rb8 {2.19/19 83} 55. Qc3 {
(Qc2) 55} Ra8 {2.19/18 48} 56. Bf4 {(Qb2) 47} Bh7 {3.20/19 77} 57. Rc7 {12} Rb8
{3.64/19 61} 58. Qc6 {11} Qe8 {3.81/19 76} 59. Qxe8+ {13} Rxe8 {3.81/19 1} 60.
Nd2 {(Rd7) 8} Bg8 {4.64/18 51} 61. Kf3 {9} g6 {5.02/21 24} 62. f6 {(fxg6) 11}
Bf8 {5.58/18 66} 63. Nc4 {10} Rd8 {5.02/14 1} 64. Rc6 {81} Ra8 {6.45/15 1} 65.
Bxd6 {77} Bxd6 {6.83/14 1} 66. Nxd6 {7} 1-0

[Event "Clash of the Computer Titans"]
[Site "?"]
[Date "2007.09.22"]
[Round "4"]
[White "Rybka"]
[Black "Zappa"]
[Result "0-1"]
[ECO "C92"]
[Annotator "Rybka"]
[PlyCount "360"]
[TimeControl "5400+20"]

{512MB, MexicoB.ctg} 1. e4 {B/0 0} e5 {11} 2. Nf3 {B/0 0} Nc6 {10} 3. Bb5 {
B/0 0} a6 {8} 4. Ba4 {B/0 0} Nf6 {9} 5. O-O {B/0 0} Be7 {8} 6. Re1 {B/0 0} b5 {
7} 7. Bb3 {B/0 0} d6 {12} 8. c3 {B/0 0} O-O {9} 9. h3 {B/0 0} Bb7 {9} 10. d4 {
B/0 0} Re8 {9} 11. Nbd2 {B/0 0} Bf8 {8} 12. a4 {B/0 0} h6 {17} 13. Bc2 {B/0 0}
exd4 {9} 14. cxd4 {B/0 0} Nb4 {11} 15. Bb1 {B/0 0} c5 {22} 16. d5 {B/0 0} Nd7 {
11} 17. Ra3 {B/0 0} c4 {11} 18. axb5 {B/0 0} axb5 {10} 19. Nd4 {B/0 0} Qb6 {11}
20. Nf5 {B/0 0} Ne5 {9} 21. Rg3 {B/0 0} g6 {11} 22. Nf3 {B/0 0} Ned3 {12} 23.
Be3 {B/0 0} Qd8 {10} 24. Bxh6 {B/0 0} Qf6 {Last book move. 15} 25. Qd2 {B/0 0}
Nxe1 {519} 26. Bxf8 {B/0 0} Nxf3+ {221} 27. Rxf3 {Last book move. B/0 0} Kxf8 {
177} 28. Nh6 {0.11/24 251} Qg7 {21} 29. Rxf7+ {0.11/23 39} Qxf7 {10} 30. Nxf7 {
0.11/25 78} Kxf7 {7} 31. Qxb4 {0.70/25 269} Ra1 {12} 32. Qxb5 {0.83/24 80}
Rxb1+ {269} 33. Kh2 {0.84/27 1} Re7 {8} 34. Qb4 {0.88/25 167} Rd7 {205} 35. Qd2
{1.00/26 402} Kg8 {166} 36. Qc2 {1.04/25 112} Ra1 {(Rf1) 64} 37. Qxc4 {
1.04/23 2} Ra8 {(Kh7) 112} 38. b4 {1.04/20 25} Rf8 {187} 39. f3 {1.12/24 14}
Rc8 {(g5) 120} 40. Qd4 {1.23/20 23} Rf7 {128} 41. h4 {1.38/24 34} Rcc7 {
(Rc2) 114} 42. b5 {1.41/23 31} Rc5 {53} 43. Qb4 {1.41/23 10} Kh7 {138} 44. b6 {
1.49/26 83} Rd7 {(Rg7) 35} 45. Qa4 {1.49/24 93} Re7 {154} 46. Kg3 {1.50/26 145}
Rc8 {9} 47. Qb4 {1.51/25 301} Rd8 {(Rd7) 12} 48. Kf4 {1.58/20 28} Bc8 {
(Kg8) 116} 49. g4 {1.65/21 26} Rf8+ {106} 50. Ke3 {1.72/25 244} Rd8 {9} 51. h5
{1.73/24 101} gxh5 {222} 52. gxh5 {1.80/25 121} Bb7 {(Kh6) 175} 53. Qb2 {
1.86/23 48} Rf8 {13} 54. Qh2 {2.00/23 7} Rf6 {(Rd7) 90} 55. Qh4 {2.00/24 20}
Ref7 {98} 56. f4 {2.00/28 50} Rh6 {(Rf8) 33} 57. Qg5 {2.00/26 142} Ba8 {
(Rhf6) 92} 58. Kd3 {2.00/25 47} Bb7 {114} 59. Kd4 {2.00/27 54} Rhf6 {100} 60.
f5 {2.00/28 34} Ba6 {(Rf8) 68} 61. Kc3 {2.00/24 27} Bb7 {50} 62. Kb4 {2.00/27 4
} Ba6 {(Rf8) 72} 63. Ka5 {2.00/24 26} Bb7 {69} 64. Kb5 {2.00/28 2} Ba8 {
(Kh8) 76} 65. Qg6+ {2.00/23 16} Kh8 {82} 66. Qg3 {2.00/27 10} Bb7 {78} 67. Qf4
{2.00/26 2} Rh7 {(Kh7) 68} 68. Qd2 {2.00/25 32} Rhh6 {(Rhf7) 64} 69. Qg5 {
2.01/25 19} Kh7 {9} 70. Kc4 {2.01/26 16} Ba6+ {(Kh8) 50} 71. Kd4 {2.01/27 19}
Bb7 {88} 72. Kc3 {2.01/29 102} Bc8 {(Ba6) 153} 73. Kc2 {2.01/24 27} Ba6 {77}
74. Kb2 {2.01/29 63} Bb7 {127} 75. Kc1 {2.01/29 123} Ba6 {(Ba8) 63} 76. Kc2 {
2.01/27 53} Bb7 {9} 77. Kb2 {2.01/26 6} Ba6 {(Ba8) 94} 78. Kb3 {2.01/27 38} Bb7
{57} 79. Kc2 {0.00/28 179} Ba6 {54} 80. Kc1 {2.01/28 152} Kh8 {(Bb7) 72} 81.
Kb1 {1.85/25 38} Kh7 {398} 82. Ka2 {1.85/31 1} Bb7 {68} 83. Kb3 {1.85/29 10}
Kh8 {(Ba6) 34} 84. Kb4 {1.85/24 17} Kh7 {(Ba8) 14} 85. Kb5 {1.85/27 104} Kh8 {
13} 86. Ka4 {1.85/26 3} Kh7 {68} 87. Ka5 {1.85/28 78} Kh8 {93} 88. Qh4 {
1.85/29 55} Rf7 {(Kg8) 72} 89. Qf4 {1.85/25 13} Kh7 {(Rhf6) 62} 90. Kb5 {
1.85/25 16} Rff6 {(Rhf6) 33} 91. Qc1 {1.85/25 15} Rf7 {10} 92. Qa3 {1.85/25 8}
Rg7 {47} 93. Qf3 {1.85/26 22} Rf6 {(Rf7) 43} 94. Qc3 {1.85/24 30} Rgf7 {
(Rf8) 47} 95. Qd2 {1.85/26 65} Rh6 {42} 96. Qg2 {1.85/27 38} Rg7 {56} 97. Qh2 {
1.85/27 38} Kg8 {(Rf6) 58} 98. Qh4 {1.85/24 12} Kh7 {22} 99. Qh1 {1.85/25 6}
Rf7 {(Kg8) 38} 100. Ka5 {1.85/23 32} Rg7 {(Kg7) 54} 101. Qh4 {1.85/25 48} Rd7 {
(Rf7) 88} 102. Qg3 {1.85/23 66} Rg7 {38} 103. Qf3 {1.85/27 41} Re7 {(Rf7) 64}
104. Kb5 {1.85/23 12} Rg7 {(Rf6) 46} 105. Qc3 {1.85/25 47} Rf7 {46} 106. Qh3 {
1.85/25 83} Rhf6 {(Re7) 40} 107. Qe3 {0.86/22 37} Rg7 {(Rh6) 55} 108. Qc1 {
0.86/23 28} Rgf7 {(Rd7) 8} 109. h6 {0.86/24 25} Rxh6 {43} 110. Qd2 {0.86/25 11}
Kg7 {(Rhf6) 42} 111. Qg5+ {0.86/24 13} Kh7 {7} 112. Kb4 {0.86/25 4} Rhf6 {
(Rg7) 42} 113. Kc4 {0.86/23 29} Ba6+ {(Rf8) 43} 114. Kb3 {0.86/23 16} Bc8 {
(Rh6) 44} 115. Kc3 {0.86/21 13} Ba6 {(Rh6) 42} 116. Kb2 {0.85/20 33} Rh6 {34}
117. Kb3 {0.81/23 18} Rg7 {34} 118. Qc1 {0.81/23 22} Re7 {(Bb7) 9} 119. Qf4 {
0.86/21 30} Rb7 {(Rf6) 31} 120. f6 {1.25/19 28} Rf7 {9} 121. Qxd6 {1.16/19 1}
Rhxf6 {37} 122. Qe5 {0.44/19 65} Rxb6+ {38} 123. Kc3 {0.39/20 18} Rh6 {
(Rfb7) 16} 124. Kd4 {0.35/19 49} Be2 {14} 125. Qe8 {0.09/20 62} Bh5 {(Rhf6) 8}
126. Qd8 {0.15/20 34} Bg4 {69} 127. e5 {0.33/20 35} Rd7 {(Be6) 41} 128. Qe8 {
0.00/20 21} Kg7 {46} 129. Qb8 {-0.06/22 13} Be6 {34} 130. d6 {-0.13/22 32} Rh4+
{11} 131. Ke3 {-0.12/22 22} Kg6 {(Bd5) 13} 132. Qf8 {-0.21/21 94} Rf7 {7} 133.
Qd8 {-0.21/21 7} Rb4 {8} 134. Qe8 {-0.22/21 23} Kf5 {8} 135. Qh8 {-0.22/20 1}
Rb3+ {48} 136. Kd4 {-0.26/22 21} Kg4 {9} 137. Qg8+ {-0.26/21 9} Kf3 {6} 138.
Qa8+ {-0.28/21 72} Kg3 {(Rfb7) 44} 139. Qg8+ {-0.21/18 9} Kf2 {7} 140. Qg5 {
-0.21/20 6} Rb4+ {9} 141. Kc5 {-0.30/21 28} Rff4 {7} 142. d7 {-0.33/21 8} Bxd7
{10} 143. Qh5 {-0.35/21 25} Rfc4+ {67} 144. Kd6 {-0.35/24 25} Rd4+ {6} 145. Ke7
{-0.35/24 85} Bg4 {8} 146. Qh2+ {-0.35/22 72} Kf3 {7} 147. e6 {-0.37/19 22} Re4
{60} 148. Qh1+ {-0.41/19 3} Kf4 {78} 149. Qc1+ {-0.43/21 1} Kf5 {6} 150. Qf1+ {
-0.43/21 20} Ke5 {11} 151. Qa1+ {-0.44/21 33} Rbd4 {52} 152. Qg1 {-0.44/22 28}
Bxe6 {81} 153. Qg5+ {-0.50/23 1} Bf5 {7} 154. Qg3+ {-0.51/23 44} Kd5+ {(Rf4) 28
} 155. Kf6 {-0.44/19 13} Rf4 {(Be6) 43} 156. Qb3+ {-0.50/16 10} Kc6 {45} 157.
Kg5 {-0.80/23 10} Bd7 {33} 158. Qc2+ {-0.90/23 1} Rc4 {117} 159. Qg2+ {
-10.22/24 1} Kc5 {40} 160. Qg1+ {-6.93/23 31} Kd5 {59} 161. Qg2+ {-10.30/24 1}
Ke6 {48} 162. Qg3 {-10.28/23 11} Rf8 {(Rf5+) 9} 163. Qh3+ {-10.07/24 105} Ke7 {
71} 164. Qe3+ {-10.38/25 18} Kf7 {44} 165. Qf2+ {-10.51/24 24} Kg8 {34} 166.
Qa2 {-10.34/23 15} Be6 {40} 167. Qa1 {-10.51/21 5} Rg4+ {(Rf5+) 36} 168. Kh6 {
-10.51/20 13} Rc8 {(Rff4) 36} 169. Qa6 {-#23/19 14} Bd7 {10} 170. Qa2+ {
-#22/17 1} Rcc4 {11} 171. Qa7 {-#21/16 1} Rcd4 {(Bc8) 49} 172. Qa2+ {-#20/12 5}
Kf8 {9} 173. Qb3 {-#19/11 1} Ke7 {(Rh4+) 9} 174. Qa3+ {-#11/11 3} Rb4 {(Kf7) 9}
175. Qe3+ {-#9/10 2} Kf6 {11} 176. Qf3+ {-#10/4 1} Rgf4 {(Rbf4) 7} 177. Qc3+ {
-#7/3 2} Rbd4 {8} 178. Qf3 {-#5/3 1} Rxf3 {14} 179. Kh7 {-#2/3 1} Bf5+ {(Be6) 7
} 180. Kh8 {-#1/3 2} Rd8# 0-1
Parent - - By turbojuice1122 (Gold) [us] Date 2007-09-23 01:59
It would be amusing if this wasn't such a serious match, but it seems that game 4 would have been a "forced draw" by operator agreement if not for the 50-move rule.
Parent - - By Alkelele (****) Date 2007-09-23 02:03
How can you force a draw by agreement if the opponent insists on keeping on playing?
Parent - By turbojuice1122 (Gold) [us] Date 2007-09-23 02:55
I think that when the time comes for the next game to roll around after the 1500th move, the operators would naturally be forced to agree to a draw.
Parent - - By Banned for Life (Gold) Date 2007-09-23 02:12
Game 3 illustrates a time management problem common with engines and the first move out of book. Whereas a person or a centaur would spend a lot of time on this critical move, an engine treats it just like any other move and ends up with reduced depth due to not pondering and not having anything in the hash table. I'm not sure whether the new version of Rybka would come to a different conclusion with greater depth, but achieving only depth 15 for this critical move while using an octal in a 60-20 game is really unacceptable.

Regards,
Alan
Parent - By XmikeX (**) [us] Date 2007-09-23 05:22
""Whereas a person or a centaur would spend a lot of time on this critical move, an engine treats it just like any other move and ends up with reduced depth due to not pondering and not having anything in the hash table. ""

-- Seems like this exception would be trivial to code into any program (?????)
Parent - - By Alkelele (****) Date 2007-09-23 08:56
It's true that more time should have been spent on this move, but, for good measure, the depth counts of the latest version are not directly comparable to the depth count of earlier versions.
Parent - - By Banned for Life (Gold) Date 2007-09-23 17:54
Dagh,

Of course any time you change the search, you also change the meaning of depth. But if you had only a single control that told the engine when to move, I am very certain that you would not have made the first move out of book in a complex position in a 60-20 game in 68 seconds. In fact, I would be much less surprised if you spent 680 seconds on this move. Obviously this problem isn't Rybka specific. Its just more annoying in Rybka because of the very high level of the remainder of her game.

Regards,
Alan
Parent - By turbojuice1122 (Gold) [us] Date 2007-09-23 18:00
In my current 60'+20" match on my computer between Rybka 2.3.2a and Zappa Mexico, both engines typically seem to take quite a long time on the first moves out of book--in fact, I had one where Rybka took even longer than your example 680 seconds :-)
Parent - By Vasik Rajlich (Silver) [hu] Date 2007-10-01 22:53
Dagh managed to convince me about exactly this in Mexico. Another argument is that typically, the book authors will (tend to) put the obvious moves in their books, so that the first positions out of book will (tend to) be more critical.

Vas
Parent - - By oudheusa (*****) [nl] Date 2007-09-23 07:45
Not the relevant 6 man tablebases?

I'm sorry to say to you guys seem pretty sloppy about this match or does this fall under the category bad luck?

The jokes after games 1 and 2 about playing with the 2.3.2a version and without tablebases 'to let Zappa get into the match' are quite telling imo.

Ok so we all expected a walk over and it is not happening.

Time to get serious, fix the book, use full tbs and play with a solid tested version.
Parent - By The Wizard (***) Date 2007-09-23 07:55
WOW!! went out to dinner last night with Rybka on the back foot in game 3 only to wake up this morning to find Rybka lost 2 games in a row!!!!  .... Is that octo it is running on working OK? lolol :)

Regards
Parent - - By Uri Blass (*****) [il] Date 2007-09-23 08:48
It is very easy to criticize other people when you are not playing.
You can be sure that the rybka team do the best in order to win the match based on their understanding.

I can only suggest to have some dynamic comtempt factor that means that if your evaluation is clearly higher than draw you should give draw a positive score so the program is not going to sacrifice too much in order to avoid draw(some sacrificing in order to avoid draw make sense but probably not if the score drops from more than +2 to less than +1)

Uri
Parent - - By M ANSARI (*****) [kw] Date 2007-09-23 09:45
I actually left early since I thought there was no way to progress ... I come back this morning and find out Rybka actually lost !!!  All I can say is that there are some SERIOUS flaws in Rybka's evaluation of endgames with pawns.  Rybka does not realize the importance of connected pawns and treats connected pawns same as pawns that are seperated.  Qe1! which wins on the spot was missed due to the fact that Rybka does not evaluate the importance of connected passed pawns.  DF10 finds the move in 6 seconds on my computer.  I will not get into how the game was lost because it simply follows the same idea that the evaluation is simply wrong.  When I first learned chess I remember a simple rule that said a Queen is worth 9 points a pawn 1 point a bishop or a knight is 3 points and a rook is 5 points.  A quick look at Rybka's evaluation shows that in this position Rybka thinks it has the edge !!!

8/5r1k/br6/3PQ3/4P3/2K5/8/8 b - - 0 123


Well let us see ... Queen + 2 pawns = 11 points .... 2 rooks + bishop is 13 points ... how on earth is this supposed to be +0.12 at depth 18 ????  White is 2 points down !!!  I would like to add that I have seen similar positions where Rybka thinks that 3 seperated pawns are even or better than a lone bishop.  Obviously a bishop can easily be stronger than the pawns and the pawns will fall in the long term since they are not supported.  I am sure that Erdo was aware of this weakness otherwise he would have accepted a quick draw.

Without being too harsh on the Rybka team (I still think Rybka will win the match) ... I think it is back to the drawing board.  Obviously there is a need of a dynamic evaluation according to the stage of the game.  The evaluation that works so impressively well in the opening and in the middle game ... can fail miserably in the endgame.  Changing the evaluation would probably make Rybka weaker generally ... but surely a dynamic evaluation that would change according to the phase of the game would have the best of both worlds.  Maybe in Rybka 3.0
Parent - - By TedSummers (***) [us] Date 2007-09-23 10:06
Hi Ansari,

"Well let us see ... Queen + 2 pawns = 11 points .... 2 rooks + bishop is 13 points ... how on earth is this supposed to be +0.12 at depth 18 ????"

I guess it gave it's self points for having control of the center and connected passed pawns as well. Not that it matters, the evaluation was a little off here. I thought the loss had more to do with the contempt factor.
Parent - By Alkelele (****) Date 2007-09-23 10:11
Yes, probably there was a bonus for some "nice" connected and fairly advanced pawns. You would have to ask Larry and Vas for details, but I can assure you that some thought went into this game during dinner afterwards... BTW, it should be said that several strong players (IM strength at least) were following the game, and most were surprised how critical this position was for black.

As far as I know, we didn't use contempt in this game (stuff like this is hard-coded in the latest versions, so I can't see as operator).
Parent - By Uri Blass (*****) [il] Date 2007-09-23 10:58
Note that rybka has connected passed pawns in the position that you posted so increasing the score for connected pawns is not a solution to the problem.

Uri
Parent - By Sesse (****) [no] Date 2007-09-23 09:05
So, a bit of bad luck, combined with very wise decisions from the Zappa team, and Rybka is down 2.5-1.5. Thankfully there's six more games to be played! This will be a very exciting match for sure, good luck in the continuation :-)

/* Steinar */
Parent - - By Nelson Hernandez (Gold) [us] Date 2007-09-23 11:16
First of all, thanks for your candid comments.  Despite the cries of anguish about your hideous performance on "Sabado Negra" I am actually glad the Rybka team went through this terrible ordeal.  I'm rooting for you, of course.  But what I root for even more are the engine improvements that will inevitably flow from this smack-down.

One cannot help but smile at Erdo's cunning.  What a guy, huh?
Parent - - By Harvey Williamson (*****) Date 2007-09-23 11:52
One cannot help but smile at Erdo's cunning.  What a guy, huh?

Hi Nelson,

Here is erdo hard at work before the tournament at the secret team Hiarcs camp:

http://www.youtube.com/watch?v=mT6xCzMAKXc&mode=related&search=

Best Wishes,

Harvey
Parent - By Nelson Hernandez (Gold) [us] Date 2007-09-23 15:00
Brutal!
Parent - - By Nelson Hernandez (Gold) [us] Date 2007-09-23 11:23
But I should mention, as I did earlier to Larry K., that while tablebases generally have only perhaps a +5 ELO value, they have +0.5 or +1.0 value when you don't have them at the critical moment!  Also, I think the +5 ELO assessment is perhaps true when looking at a vast body of games in a database.  But surely EGTBs are worth a LOT more in a match like this, which is VERY OFTEN decided in the endgame.  We're talking about two superb books that will typically result in a close game, two closely matched engines...it is just inevitable that you will get to the endgame often with the game in the balance.

You MUST have complete EGTBs for a tournament game like this.  The trick is getting them all loaded beforehand.
Parent - - By Uri Blass (*****) [il] Date 2007-09-23 11:32
Note that the question is not how often tablebase positions happen in games but how often the engine goes wrong because of not having tablebases.

We have no proof that tablebases could change the result of the existing games.
tablebases could certainly help rybka to see earlier that it lose the game but I doubt if they could prevent her to lose the game.

Uri
Parent - By George Tsavdaris (****) Date 2007-09-23 11:42

>Note that the question is not how often tablebase positions happen in games but how often the engine goes wrong because of not having >tablebases.


Yes, although i think more critical is the following:
If someone proves(via tests of many number of games but also small number of games) that tablebases do NOT hurt performance, or with other words the gain of ELO is zero or more than zero(even very small), then having tablebases is mandatory!

Since if you don't lose, but you only can win, even if most times you will not win, then it's logical to use tbs....
Parent - - By George Tsavdaris (****) Date 2007-09-23 11:38

>But I should mention, as I did earlier to Larry K., that while tablebases generally have only perhaps a +5 ELO value, they have +0.5 or +1.0 value >when you don't have them at the critical moment!  Also, I think the +5 ELO assessment is perhaps true when looking at a vast body of games in a >database.  But surely EGTBs are worth a LOT more in a match like this, which is VERY OFTEN decided in the endgame.


Exactly! You should tell them that once again.... :)

When you play 800 CEGT games, well maybe +5 ELO.
When you play such small number of games like now, the gains increase dramatically if you are lucky enough to reach such a position. It can give you 0.5 or 1 full point!
If you are not lucky enough then no problem! No decrease of strength. Just like not having the tbs. But what happens if you are lucky?

Verdict: You've got to have tablebases!
Parent - By TedSummers (***) [us] Date 2007-09-23 11:45
Yes, Well said and point worth taking George.
Up Topic Rybka Support & Discussion / Rybka Discussion / PGNs of game 3 and 4 with Rybka evaluation, Zappa won both

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill