Topic Rybka Support & Discussion / Rybka Discussion / PGN of game 5 with Rybka eval, Zappa won third game in a row
Rybka got an inferior position after f7-f5. It's a curiosity that Joel Benjamin in his match vs. Rybka criticised the same move harshly (without being able to punish it). Larry says that the move might be motivated by the wish to decentralise white's night on e4, but of course, the move left Rybka with a weak pawn on e6 and a good e5-square for Zappa. Rybka was also probably partly punished for her not so sophisticated time management in that she spent a lot of time finishing iterations on pondered moves, burning lots of minutes in the shuffling phase in the middlegame (the latest version should have a much better time management, but today we played it safe with the intermediate version, hehe :-)). Eventually, 75.Qe8! decided the game. Rybka had evaluated 75.c4 down to an 0.45 score, and after about one minute of search, Zappa also had a drop from 1.8 to 1.1 or so with c4. Then Zappa found Qe8+. Congrats. Now, we REALLY have to be creative :-)
[Event "Clash of the Computer Titans"]
[Site "?"]
[Date "2007.09.23"]
[Round "5"]
[White "Zappa"]
[Black "Rybka"]
[Result "1-0"]
[ECO "B52"]
[Annotator "Rybka"]
[PlyCount "258"]
[TimeControl "4800+20"]
{512MB, MexicoB.ctg} 1. e4 {0} c5 {B/0 0} 2. Nf3 {10} d6 {B/0 0} 3. Bb5+ {11}
Bd7 {B/0 0} 4. Bxd7+ {9} Qxd7 {B/0 0} 5. O-O {15} Nf6 {B/0 0} 6. e5 {17} dxe5 {
B/0 0} 7. Nxe5 {10} Qc8 {B/0 0} 8. Nc3 {10} Nc6 {0.12/20 171} 9. Re1 {(f4) 135}
Nxe5 {0.06/19 71} 10. Rxe5 {10} e6 {0.08/20 157} 11. d3 {(Qf3) 9} Qc6 {
0.07/20 73} 12. Re1 {124} Be7 {0.06/20 61} 13. b3 {120} O-O {0.08/21 414} 14.
Bb2 {9} Rad8 {0.07/20 43} 15. Qe2 {(a4) 182} Nd7 {0.03/20 96} 16. Ne4 {(h3) 307
} f5 {0.15/20 132} 17. Nd2 {77} Bf6 {0.13/20 1} 18. Bxf6 {130} Rxf6 {0.13/20 40
} 19. Nf3 {91} Qc7 {0.19/20 350} 20. Qd2 {(a4) 78} Nb8 {0.13/20 198} 21. Re3 {
71} Nc6 {0.16/20 2} 22. Rae1 {81} Qd6 {0.19/20 102} 23. Qe2 {(Ne5) 16} Re8 {
0.19/20 68} 24. Ng5 {(Ne5) 23} Rh6 {0.16/18 18} 25. h3 {82} Nd4 {0.18/21 15}
26. Qd1 {13} Qd8 {0.20/20 99} 27. Nf3 {13} Nc6 {0.20/20 113} 28. Ne5 {80} Qc7 {
0.21/20 5} 29. Qc1 {(Qd2) 79} Rf6 {0.20/19 70} 30. Nxc6 {(f4) 105} Qxc6 {
0.21/19 22} 31. Re5 {29} Rg6 {0.23/20 42} 32. g3 {12} Rh6 {0.24/20 42} 33. Kh2
{(h4) 68} b6 {0.26/20 48} 34. Qe3 {11} Rf6 {0.31/21 240} 35. Re2 {(a4) 84} Qd6
{0.27/19 151} 36. a4 {(a3) 72} h6 {0.34/18 54} 37. a5 {(f4) 98} g5 {0.23/18 102
} 38. Qf3 {(Kg2) 65} Re7 {0.22/20 51} 39. Kg1 {(Re1) 14} Kf8 {0.22/18 45} 40.
Re1 {100} Kg7 {0.23/22 69} 41. axb6 {(Kf1) 81} axb6 {0.23/20 14} 42. Qa8 {140}
Rf8 {0.23/22 51} 43. Qa1 {71} Rf6 {0.24/22 114} 44. R1e2 {(Qc3) 22} Kf7 {
0.25/19 47} 45. Qa8 {(Qa4) 98} Qd7 {0.23/18 28} 46. Re1 {33} Re8 {0.21/19 49}
47. Qf3 {(Qa6) 68} Kg7 {0.17/19 17} 48. Kf1 {(Kh1) 118} Qc8 {0.18/18 10} 49.
Kg1 {(Kg2) 61} Qd7 {0.00/25 15} 50. Kf1 {(Rxc5) 55} Qc8 {0.00/24 28} 51. Ra1 {
(Rxc5) 47} Rf7 {0.19/17 39} 52. Kg1 {(Qe2) 66} Qd7 {0.18/16 10} 53. Rb1 {
(Qe2) 44} Qd6 {0.18/20 42} 54. Rbe1 {110} Ree7 {0.17/21 39} 55. Qa8 {(Kf1) 20}
Rf8 {0.00/22 10} 56. Qa6 {(Rxc5) 33} Rd8 {0.24/17 25} 57. Qa1 {(f3) 56} Kg6 {
0.21/15 10} 58. R1e2 {27} Kf7 {0.32/19 68} 59. R2e3 {(Qa4) 15} Rf8 {0.28/19 60}
60. Qa4 {(Re1) 10} Kf6 {0.19/17 32} 61. Qa6 {(f4) 71} f4 {0.45/19 108} 62. gxf4
{68} gxf4 {0.34/22 21} 63. Re2 {(Re1) 33} Rg8+ {0.24/20 26} 64. Kf1 {6} Rg5 {
0.26/20 5} 65. Qc4 {6} Rf5 {0.35/20 16} 66. Ke1 {(b4) 25} Qxe5 {0.42/19 26} 67.
Rxe5 {10} Rxe5+ {0.41/20 14} 68. Kf1 {30} Kg5 {0.38/21 28} 69. Qa6 {88} f3 {
0.57/24 3} 70. Qxb6 {54} Rf7 {0.84/24 14} 71. Qd8+ {58} Kg6 {0.68/25 15} 72.
Qg8+ {7} Rg7 {0.62/25 35} 73. Qb8 {7} Rf5 {0.84/24 86} 74. d4 {80} Kh5 {
0.76/24 1} 75. Qe8+ {(c4) 131} Rg6 {0.88/21 9} 76. dxc5 {84} Rxc5 {1.39/25 2}
77. c4 {7} Rcg5 {1.45/23 22} 78. Ke1 {52} Re5+ {1.61/23 37} 79. Kd2 {7} Re2+ {
1.54/23 19} 80. Kc3 {(Kd3) 6} Rxf2 {1.62/22 21} 81. Qf7 {7} Rfg2 {2.06/23 46}
82. Qxf3+ {7} Kh4 {2.06/23 4} 83. Kb4 {108} R2g3 {2.06/24 1} 84. Qb7 {(Qe2) 49}
h5 {2.48/20 76} 85. c5 {58} Rg7 {2.90/22 6} 86. Qb6 {29} Re7 {3.22/22 38} 87.
c6 {45} Rgg7 {3.60/22 19} 88. Qe3 {58} e5 {3.67/22 5} 89. Kc5 {227} e4 {
4.85/22 1} 90. Kd6 {(b4) 8} Ref7 {4.60/21 71} 91. b4 {27} Rg6+ {5.08/20 33} 92.
Ke5 {7} Rg5+ {5.09/22 51} 93. Ke6 {36} Rfg7 {5.09/21 16} 94. Qxe4+ {9} Kxh3 {
5.09/20 43} 95. Kf6 {(c7) 8} Rg4 {5.25/14 29} 96. Qh1+ {(Qc2) 54} Kg3 {
5.09/16 2} 97. Qxh5 {7} Rg8 {5.09/13 27} 98. Qxg4+ {(b5) 11} Rxg4 {#37/8 5} 99.
b5 {10} Rc4 {0.01/3 1} 100. Ke5 {8} Rc5+ {0.01/3 1} 101. Kd6 {7} Rxb5 {0.01/3 1
} 102. c7 {6} Rb6+ {0.01/3 1} 103. Kd5 {7} Rb5+ {0.01/3 1} 104. Kd4 {7} Rb4+ {
0.01/3 1} 105. Kc3 {7} Rb1 {0.01/3 1} 106. Kc2 {8} Rf1 {0.01/3 1} 107. c8=Q {10
} Rf3 {0.01/3 1} 108. Kd2 {(Qe6) 12} Kf4 {#27/3 2} 109. Qc6 {(Qe6) 9} Rf2+ {
#26/3 2} 110. Kd3 {7} Kg5 {#25/3 1} 111. Qc1+ {(Qc8) 8} Rf4 {#24/3 2} 112. Ke3
{(Qc5+) 9} Rf6 {0.01/3 2} 113. Ke4+ {9} Kg6 {0.01/3 1} 114. Qc3 {(Qc8) 10} Re6+
{#21/3 2} 115. Kd5 {8} Re8 {#20/3 1} 116. Qc6+ {13} Kf7 {#19/3 1} 117. Qd7+ {7}
Re7 {0.01/3 1} 118. Qf5+ {7} Kg7 {0.01/3 1} 119. Kd6 {7} Ra7 {#16/3 1} 120.
Qe5+ {6} Kh7 {#15/3 1} 121. Ke6 {10} Ra2 {0.01/3 1} 122. Qc7+ {(Qe4+) 9} Kg6 {
#13/3 2} 123. Qc4 {7} Ra5 {#12/3 1} 124. Qd3+ {8} Kh5 {0.01/3 1} 125. Kf6 {8}
Kh4 {#10/3 1} 126. Qc4+ {8} Kh5 {#9/3 1} 127. Qe2+ {(Qc3) 8} Kh4 {#8/3 2} 128.
Qe1+ {7} Kg4 {#7/3 1} 129. Qxa5 {19} Kf3 {#6/3 1} 1-0
[Event "Clash of the Computer Titans"]
[Site "?"]
[Date "2007.09.23"]
[Round "5"]
[White "Zappa"]
[Black "Rybka"]
[Result "1-0"]
[ECO "B52"]
[Annotator "Rybka"]
[PlyCount "258"]
[TimeControl "4800+20"]
{512MB, MexicoB.ctg} 1. e4 {0} c5 {B/0 0} 2. Nf3 {10} d6 {B/0 0} 3. Bb5+ {11}
Bd7 {B/0 0} 4. Bxd7+ {9} Qxd7 {B/0 0} 5. O-O {15} Nf6 {B/0 0} 6. e5 {17} dxe5 {
B/0 0} 7. Nxe5 {10} Qc8 {B/0 0} 8. Nc3 {10} Nc6 {0.12/20 171} 9. Re1 {(f4) 135}
Nxe5 {0.06/19 71} 10. Rxe5 {10} e6 {0.08/20 157} 11. d3 {(Qf3) 9} Qc6 {
0.07/20 73} 12. Re1 {124} Be7 {0.06/20 61} 13. b3 {120} O-O {0.08/21 414} 14.
Bb2 {9} Rad8 {0.07/20 43} 15. Qe2 {(a4) 182} Nd7 {0.03/20 96} 16. Ne4 {(h3) 307
} f5 {0.15/20 132} 17. Nd2 {77} Bf6 {0.13/20 1} 18. Bxf6 {130} Rxf6 {0.13/20 40
} 19. Nf3 {91} Qc7 {0.19/20 350} 20. Qd2 {(a4) 78} Nb8 {0.13/20 198} 21. Re3 {
71} Nc6 {0.16/20 2} 22. Rae1 {81} Qd6 {0.19/20 102} 23. Qe2 {(Ne5) 16} Re8 {
0.19/20 68} 24. Ng5 {(Ne5) 23} Rh6 {0.16/18 18} 25. h3 {82} Nd4 {0.18/21 15}
26. Qd1 {13} Qd8 {0.20/20 99} 27. Nf3 {13} Nc6 {0.20/20 113} 28. Ne5 {80} Qc7 {
0.21/20 5} 29. Qc1 {(Qd2) 79} Rf6 {0.20/19 70} 30. Nxc6 {(f4) 105} Qxc6 {
0.21/19 22} 31. Re5 {29} Rg6 {0.23/20 42} 32. g3 {12} Rh6 {0.24/20 42} 33. Kh2
{(h4) 68} b6 {0.26/20 48} 34. Qe3 {11} Rf6 {0.31/21 240} 35. Re2 {(a4) 84} Qd6
{0.27/19 151} 36. a4 {(a3) 72} h6 {0.34/18 54} 37. a5 {(f4) 98} g5 {0.23/18 102
} 38. Qf3 {(Kg2) 65} Re7 {0.22/20 51} 39. Kg1 {(Re1) 14} Kf8 {0.22/18 45} 40.
Re1 {100} Kg7 {0.23/22 69} 41. axb6 {(Kf1) 81} axb6 {0.23/20 14} 42. Qa8 {140}
Rf8 {0.23/22 51} 43. Qa1 {71} Rf6 {0.24/22 114} 44. R1e2 {(Qc3) 22} Kf7 {
0.25/19 47} 45. Qa8 {(Qa4) 98} Qd7 {0.23/18 28} 46. Re1 {33} Re8 {0.21/19 49}
47. Qf3 {(Qa6) 68} Kg7 {0.17/19 17} 48. Kf1 {(Kh1) 118} Qc8 {0.18/18 10} 49.
Kg1 {(Kg2) 61} Qd7 {0.00/25 15} 50. Kf1 {(Rxc5) 55} Qc8 {0.00/24 28} 51. Ra1 {
(Rxc5) 47} Rf7 {0.19/17 39} 52. Kg1 {(Qe2) 66} Qd7 {0.18/16 10} 53. Rb1 {
(Qe2) 44} Qd6 {0.18/20 42} 54. Rbe1 {110} Ree7 {0.17/21 39} 55. Qa8 {(Kf1) 20}
Rf8 {0.00/22 10} 56. Qa6 {(Rxc5) 33} Rd8 {0.24/17 25} 57. Qa1 {(f3) 56} Kg6 {
0.21/15 10} 58. R1e2 {27} Kf7 {0.32/19 68} 59. R2e3 {(Qa4) 15} Rf8 {0.28/19 60}
60. Qa4 {(Re1) 10} Kf6 {0.19/17 32} 61. Qa6 {(f4) 71} f4 {0.45/19 108} 62. gxf4
{68} gxf4 {0.34/22 21} 63. Re2 {(Re1) 33} Rg8+ {0.24/20 26} 64. Kf1 {6} Rg5 {
0.26/20 5} 65. Qc4 {6} Rf5 {0.35/20 16} 66. Ke1 {(b4) 25} Qxe5 {0.42/19 26} 67.
Rxe5 {10} Rxe5+ {0.41/20 14} 68. Kf1 {30} Kg5 {0.38/21 28} 69. Qa6 {88} f3 {
0.57/24 3} 70. Qxb6 {54} Rf7 {0.84/24 14} 71. Qd8+ {58} Kg6 {0.68/25 15} 72.
Qg8+ {7} Rg7 {0.62/25 35} 73. Qb8 {7} Rf5 {0.84/24 86} 74. d4 {80} Kh5 {
0.76/24 1} 75. Qe8+ {(c4) 131} Rg6 {0.88/21 9} 76. dxc5 {84} Rxc5 {1.39/25 2}
77. c4 {7} Rcg5 {1.45/23 22} 78. Ke1 {52} Re5+ {1.61/23 37} 79. Kd2 {7} Re2+ {
1.54/23 19} 80. Kc3 {(Kd3) 6} Rxf2 {1.62/22 21} 81. Qf7 {7} Rfg2 {2.06/23 46}
82. Qxf3+ {7} Kh4 {2.06/23 4} 83. Kb4 {108} R2g3 {2.06/24 1} 84. Qb7 {(Qe2) 49}
h5 {2.48/20 76} 85. c5 {58} Rg7 {2.90/22 6} 86. Qb6 {29} Re7 {3.22/22 38} 87.
c6 {45} Rgg7 {3.60/22 19} 88. Qe3 {58} e5 {3.67/22 5} 89. Kc5 {227} e4 {
4.85/22 1} 90. Kd6 {(b4) 8} Ref7 {4.60/21 71} 91. b4 {27} Rg6+ {5.08/20 33} 92.
Ke5 {7} Rg5+ {5.09/22 51} 93. Ke6 {36} Rfg7 {5.09/21 16} 94. Qxe4+ {9} Kxh3 {
5.09/20 43} 95. Kf6 {(c7) 8} Rg4 {5.25/14 29} 96. Qh1+ {(Qc2) 54} Kg3 {
5.09/16 2} 97. Qxh5 {7} Rg8 {5.09/13 27} 98. Qxg4+ {(b5) 11} Rxg4 {#37/8 5} 99.
b5 {10} Rc4 {0.01/3 1} 100. Ke5 {8} Rc5+ {0.01/3 1} 101. Kd6 {7} Rxb5 {0.01/3 1
} 102. c7 {6} Rb6+ {0.01/3 1} 103. Kd5 {7} Rb5+ {0.01/3 1} 104. Kd4 {7} Rb4+ {
0.01/3 1} 105. Kc3 {7} Rb1 {0.01/3 1} 106. Kc2 {8} Rf1 {0.01/3 1} 107. c8=Q {10
} Rf3 {0.01/3 1} 108. Kd2 {(Qe6) 12} Kf4 {#27/3 2} 109. Qc6 {(Qe6) 9} Rf2+ {
#26/3 2} 110. Kd3 {7} Kg5 {#25/3 1} 111. Qc1+ {(Qc8) 8} Rf4 {#24/3 2} 112. Ke3
{(Qc5+) 9} Rf6 {0.01/3 2} 113. Ke4+ {9} Kg6 {0.01/3 1} 114. Qc3 {(Qc8) 10} Re6+
{#21/3 2} 115. Kd5 {8} Re8 {#20/3 1} 116. Qc6+ {13} Kf7 {#19/3 1} 117. Qd7+ {7}
Re7 {0.01/3 1} 118. Qf5+ {7} Kg7 {0.01/3 1} 119. Kd6 {7} Ra7 {#16/3 1} 120.
Qe5+ {6} Kh7 {#15/3 1} 121. Ke6 {10} Ra2 {0.01/3 1} 122. Qc7+ {(Qe4+) 9} Kg6 {
#13/3 2} 123. Qc4 {7} Ra5 {#12/3 1} 124. Qd3+ {8} Kh5 {0.01/3 1} 125. Kf6 {8}
Kh4 {#10/3 1} 126. Qc4+ {8} Kh5 {#9/3 1} 127. Qe2+ {(Qc3) 8} Kh4 {#8/3 2} 128.
Qe1+ {7} Kg4 {#7/3 1} 129. Qxa5 {19} Kf3 {#6/3 1} 1-0
![[us]](/mwf/v12/flags/us.png "United States")
This is indeed starting to look like DARK TIMES for the Rybka camp.Zappa is matching toe to toe ONLY program to do so thus far.Can Rybka come back,something is WRONG!!
I'm not sure what you mean by Zappa matching "toe to toe". Watching the games, Zappa appears to be clearly the superior program. Perhaps the Rybka team should take chances with the latest version now that the match is, for all practical purposes, in a "do-or-die" situation--after all, the only real problem with the latest version was that it entered a bad book line and played the first couple of moves out of book poorly due possibly to poor time management.
when R#6 will start?
I would assume it starts at 10:30 Mexico City time (which is 5 hours behind Greenwich mean time, though I don't know how that translates into Polish time) tomorrow, which is the time that the other games in the match start.
Must echo Turbo's remark. If this were a heavyweight championship fight, the champ would be slumped in his corner, battered and bleeding, talking incoherently, with his seconds frantically trying to close cuts and remind him where he is. (You there, get me the smelling salt! Keep waving that towel!)
The match is only half-over. But I have to say, looking at the pre-match press conference today for the first time...it is interesting to compare Erdo's soft-spoken, understated, deferential attitude to the terrific confidence of the Rybka team's. Chok iyi, Erdo!
Vas, forget this East European thing. You are an American through and through, with a Michigan accent no less!
The match is only half-over. But I have to say, looking at the pre-match press conference today for the first time...it is interesting to compare Erdo's soft-spoken, understated, deferential attitude to the terrific confidence of the Rybka team's. Chok iyi, Erdo!
Vas, forget this East European thing. You are an American through and through, with a Michigan accent no less!
![[ph]](/mwf/v12/flags/ph.png "Philippines")
The match situation somehow reminds me of Fischer-Spassky 1972. Boris got off to a 1.5-0.5 start, then Fischer came back strongly and won 3 of the next 4 games to impose his will on the match outcome.
Yes, the Rybka team needs to be creative.
Rybka needs to play like Topalov, who is known to make a strong comeback after a dismal start in a tournament.
Yes, the Rybka team needs to be creative.
Rybka needs to play like Topalov, who is known to make a strong comeback after a dismal start in a tournament.
![[gb]](/mwf/v12/flags/gb.png "United Kingdom")
am I right in saying that the latest version of rybka is NOT being used becuase of a bug? If so I assume that 2.3.2a is being used which is weird becuase on the ratings 2.3.2a is killing zappa mexico.
therefore could it be possible that "Zappa Mexeco" and one actually in mexeco are different, perhaps greatly so?
therefore could it be possible that "Zappa Mexeco" and one actually in mexeco are different, perhaps greatly so?
There are three versions:
1) 2.3.2a
2) A version with eval changes.
3) A version with both eval and search changes.
We played with the middle one in game 2, 4 and 5, with 2.3.2a in the first game, and we hope to play with the last one in the remaining 5 games (as we did in game 3).
I can't give the precise guesstimates of the improvement, but it's supposed to be something like 25 and 50 Elo.
1) 2.3.2a
2) A version with eval changes.
3) A version with both eval and search changes.
We played with the middle one in game 2, 4 and 5, with 2.3.2a in the first game, and we hope to play with the last one in the remaining 5 games (as we did in game 3).
I can't give the precise guesstimates of the improvement, but it's supposed to be something like 25 and 50 Elo.
This "improvement" is according to who and how was it measured? If it's an improvement then why were other versions used?
This is very interesting. If I had an 8 way box I might get interested in computer chess again.
This is very interesting. If I had an 8 way box I might get interested in computer chess again.
The version with the eval changes only is pretty well tested and I am confident that the improvement is in or close to the 20 to 30 Elo range. There is very little chance that it has any bugs other than those if any present in 2.3.2a. The version with search changes did have some problems which have supposedly been fixed, and as soon as we have enough test results to confirm that it does not have important bugs we will use it, probably tomorrow. The additional rating gain for this version is much less clear, but it should be substantial on an 8 core system if all is working right.
Dagh, I am beside myself. You guys are too clever by half! I daresay the Rybka Mexico team must have an average IQ in the 160-170 range but you guys have thoroughly outsmarted yourselves!
Not happy with one or even two versions, you go to Mexico with three. Not satisfied with a full EGTB set, or even a partial one, you have a third option, no endgame tablebases! I suppose you have several books down there? Yeah, sure, of course! How could it be otherwise?
It's clear you guys need your mojo back. But you have that in at least three or four versions too!
Not happy with one or even two versions, you go to Mexico with three. Not satisfied with a full EGTB set, or even a partial one, you have a third option, no endgame tablebases! I suppose you have several books down there? Yeah, sure, of course! How could it be otherwise?
It's clear you guys need your mojo back. But you have that in at least three or four versions too!
![[nl]](/mwf/v12/flags/nl.png "Netherlands")
Funny read :)
Sad thing is it's all true :(
Sad thing is it's all true :(
we hope to play with the (version with both eval and search changes) in the remaining 5 games.and so, the public also hopes to see better results for Rybka: 4-1 in order to win the match
When you say "on the ratings 2.3.2a is killing zappa mexico", I assume that you're referring to the CEGT and CCRL rating lists? These are really not good at all in predicting the outcome of a match like this, even if the match was long.
First of all, those rating lists use "general" books, i.e. when someone makes an engine tournament, the same opening book is used for all engines. This has a very significant random effect, and the statistics show this. Thus, the CCRL and CEGT lists are mainly good for telling you where the engines rank in their ability to correctly analyze many positions, "on average", in a wide variety of game situations. On the other hand, the SSDF list has the engines use their own books. This, therefore, will give at least some indication of how the engines will tend to fare against each other in actual match conditions, where they will have opening books specialized for that engine's strengths and that tend to avoid that engine's weaknesses. For many engines, this doesn't really make much of a difference--but for others, it makes a noticeable difference; Junior and Hiarcs are notable examples. There are definitely many types of positions that Zappa Mexico plays quite poorly--but these are easily avoided with a good specialized opening book (such opening books don't help all engines, though). This is one reason why Zappa Mexico scores much worse on the CCRL and CEGT rating lists against Rybka 2.3.2a compared with results from matches on Playchess, even in long "semiformal" matches where the hardware is basically regulated or where one can calculate the expected rating difference from the hardware difference.
Another problem is, as many have already noted, that Zappa Mexico scales better to 8 CPUs than Rybka does. However, I doubt that this difference worth is more than 20 elo, and I think this is a generous estimate.
I actually think the 8 CPU argument is possibly largely unnecessary. Watching the engine match games on my machine, Zappa Mexico thus far simply appears to be playing better chess--it looks like a stronger engine. I don't have an 8 CPU machine--I only have one CPU--but being that one of Zappa's main strengths is its scaling to both 64-bit hardware and multiple CPUs, then if my conclusion holds after more games, there will be an increase in strength difference with more CPUs.
First of all, those rating lists use "general" books, i.e. when someone makes an engine tournament, the same opening book is used for all engines. This has a very significant random effect, and the statistics show this. Thus, the CCRL and CEGT lists are mainly good for telling you where the engines rank in their ability to correctly analyze many positions, "on average", in a wide variety of game situations. On the other hand, the SSDF list has the engines use their own books. This, therefore, will give at least some indication of how the engines will tend to fare against each other in actual match conditions, where they will have opening books specialized for that engine's strengths and that tend to avoid that engine's weaknesses. For many engines, this doesn't really make much of a difference--but for others, it makes a noticeable difference; Junior and Hiarcs are notable examples. There are definitely many types of positions that Zappa Mexico plays quite poorly--but these are easily avoided with a good specialized opening book (such opening books don't help all engines, though). This is one reason why Zappa Mexico scores much worse on the CCRL and CEGT rating lists against Rybka 2.3.2a compared with results from matches on Playchess, even in long "semiformal" matches where the hardware is basically regulated or where one can calculate the expected rating difference from the hardware difference.
Another problem is, as many have already noted, that Zappa Mexico scales better to 8 CPUs than Rybka does. However, I doubt that this difference worth is more than 20 elo, and I think this is a generous estimate.
I actually think the 8 CPU argument is possibly largely unnecessary. Watching the engine match games on my machine, Zappa Mexico thus far simply appears to be playing better chess--it looks like a stronger engine. I don't have an 8 CPU machine--I only have one CPU--but being that one of Zappa's main strengths is its scaling to both 64-bit hardware and multiple CPUs, then if my conclusion holds after more games, there will be an increase in strength difference with more CPUs.
Actually, Spassky won both the first and the second games of the match, the second when Fischer defaulted.
but of course. my bad.
But at the time Fischer was simply a lot better than Spasky. (And everyone else too)
![[de]](/mwf/v12/flags/de.png "Germany")
Well, if you would allow a somewhat sarcastic comment I'd say that if Vas would lose the match, such an outcome would be good for my correspondence games, because then everyone would run and buy Zappa (it's even starting now) and analyze the games with it... ;->
Topic Rybka Support & Discussion / Rybka Discussion / PGN of game 5 with Rybka eval, Zappa won third game in a row
Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill