Author: Drexel,Michael
Date: 23:44:05 09/19/03
Go up one level in this thread
On September 19, 2003 at 20:25:15, Vincent Diepeveen wrote: >On September 19, 2003 at 19:14:00, Drexel,Michael wrote: > >>Some people here have a naive idea of what a Supercomputer can do. > >>To search a fairly balanced opening position very DIEP is absurd. > >Not really. Just watch the correspondence rating list and how many kilometers >away from here they come and you'll hell sure start to realize something. > >>Any correspondence master will come up with much better results in the same >>time. > >That only happens when the strategy of the computer is the wrong one. Which in >most positions won't happen in which case it is way stronger. In case of diep >there is a good positional evaluation then which is very crushing at big depths >corr game after corr game. Why not ask Jeroen Noomen. Last ICT3 he had a corr >position where no program would have the positional knowledge to do something. > >He tried diep and it found it after a while :) > >Good knowledge can do miracles. Hopefully You need a miracle to win the Wch. > >In tournaments what counts is the boring openings which get played by the Fritz >& co kure books which lead to very boring positions where computers can just >shuffle with pieces in far endgames. So that shows both Kure's talent as well as >how poor these programs perform in middlegame :) The Kure opening book fairy tale is boring. > >Wasting expensive cpu time to just far already solved endgame is bloody nonsense >all you need here probably is a pc with shredder and egtbs at a fast SCSI >harddisk. You are most probably wrong here. > >Note egtb's at the machine are in RAM. So you can guess what happens here just >as well as i do. When it can exchange to a won endgame it will take it, but that >counts for all programs. In this position capturing a pawn at g6 is >peanuts to evaluate by a 5 men trivially. So there's really nothing here >interesting for pc programs to find IMHO. > >Just get black into a position where you can exchange to a winning 5 men. >whether that takes 10 moves or 50 moves is not really interesting i would >argue. > > > >>Would 500 cpu's Diep be able to solve (or at least to win) this position: >> >>[D] 5k2/4R3/2K3p1/4BbPn/8/8/8/8 w - - 0 89 >> >>I assume it would shuffle around the white pieces for another 50 moves. >> >>[Event "World Championship 35th-KK5"] >>[Site "Lyon/New York"] >>[Date "1990.12.01"] >>[Round "16"] >>[White "Kasparov, Garry"] >>[Black "Karpov, Anatoly"] >>[Result "1-0"] >>[ECO "C45"] >>[WhiteElo "2800"] >>[BlackElo "2730"] >>[PlyCount "203"] >>[EventDate "1990.10.??"] >>[Source "?"] >> >>1. e4 e5 2. Nf3 Nc6 3. d4 exd4 4. Nxd4 Nf6 5. Nxc6 bxc6 6. e5 Qe7 7. Qe2 Nd5 8. >>c4 Nb6 9. Nd2 Qe6 10. b3 a5 11. Bb2 Bb4 12. a3 Bxd2+ 13. Qxd2 d5 14. cxd5 cxd5 >>15. Rc1 O-O 16. Rxc7 Qg6 17. f3 Bf5 18. g4 Bb1 19. Bb5 Rac8 20. Rxc8 Rxc8 21. >>O-O h5 22. h3 hxg4 23. hxg4 Bc2 24. Qd4 Qe6 25. Rf2 Rc7 26. Rh2 Nd7 27. b4 axb4 >>28. axb4 Nf8 29. Bf1 Bb3 30. Bd3 Bc4 31. Bf5 Qe7 32. Qd2 Rc6 33. Bd4 Ra6 34. >>Bb1 Ra3 35. Rh3 Rb3 36. Bc2 Qxb4 37. Qf2 Ng6 38. e6 Rb1+ 39. Bxb1 Qxb1+ 40. Kh2 >>fxe6 41. Qb2 Qxb2+ 42. Bxb2 Nf4 43. Rh4 Nd3 44. Bc3 e5 45. Kg3 d4 46. Bd2 Bd5 >>47. Rh5 Kf7 48. Ba5 Ke6 49. Rh8 Nb2 50. Re8+ Kd6 51. Bb4+ Kc6 52. Rc8+ Kd7 53. >>Rc5 Ke6 54. Rc7 g6 55. Re7+ Kf6 56. Rd7 Ba2 57. Ra7 Bc4 58. Ba5 Bd3 59. f4 >>exf4+ 60. Kxf4 Bc2 61. Ra6+ Kf7 62. Ke5 Nd3+ 63. Kxd4 Nf2 64. g5 Bf5 65. Bd2 >>Ke7 66. Kd5 Ne4 67. Ra7+ Ke8 68. Be3 Nc3+ 69. Ke5 Kd8 70. Bb6+ Ke8 71. Rc7 Ne4 >>72. Be3 Ng3 73. Bf4 Nh5 74. Ra7 Kf8 75. Bh2 Ng7 76. Bg1 Nh5 77. Bc5+ Kg8 78. >>Kd6 Kf8 79. Bd4 Bg4 80. Be5 Bf5 81. Rh7 Kg8 82. Rc7 Kf8 83. Kc6 Kg8 84. Re7 Kf8 >>85. Bd6 Kg8 86. Re8+ Kf7 87. Re7+ Kg8 88. Be5 Kf8 89. Ra7 Bg4 90. Kd6 Bh3 91. >>Ra3 Bg4 92. Re3 Bf5 93. Kc7 Kf7 94. Kd8 Bg4 95. Bb2 Be6 96. Bc3 Bf5 97. Re7+ >>Kf8 98. Be5 Bd3 99. Ra7 Be4 100. Rc7 Bb1 101. Bd6+ Kg8 102. Ke7 1-0 >> >>Michael
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.