Author: Robert Hyatt
Date: 09:55:14 12/03/99
Go up one level in this thread
On December 03, 1999 at 12:18:29, Christophe Theron wrote: >On December 03, 1999 at 11:20:49, Robert Hyatt wrote: > >>On December 02, 1999 at 23:21:22, Chessfun wrote: >> >>> >>>This game was one of four almost identical wins by shutka. >>>All four wins were 3/0 and in most games shutka had 2 mins or so left. >>>Shutka has played almost 4000 blitz games rating is 2045. prior to this match-up >>>it was 1990. >>>Chesspartner is chesstiger 12.0e. >>> >>> >> >>I looked at the game, and I have watched this player. All that Tiger needs >>is some code to prevent blocking the position up so badly... > > >Actually Tiger has had this kind of problem for years against the human players >here. So I have added a "anti-human" code that prevents blocked positions. I >have no idea if it weakens the program against other programs. > >It works rather well, except that it is not turned ON on the computer currently >playing at FICS... I suppose we should turn it on, as there are mainly human >players there... > > the only warning I'll give is that how you define 'blocked' is critical. In the Levy match in 1984 where Levy killed us badly (Cray Blitz) Ivan Bratko (yes, the one from the bratko-kopec test sets) suggested that we simply count pawn 'rams' and penalize for them. That isn't enough. IE if you both have a pawn on the a-file, and no pawn on the b-file, the a-pawns are blocked whether they are in contact or not. Ditto for structures like d4-e3-f4 for white and d5-e6-f5 for black. There are three blocked pawns there, not 2. I won't ask how you detect blocked positions... I do it differently now and count viable pawn levers instead, which works pretty well. I then add in the actually blocked pawns for a bigger negative incentive... > >>For the last week, 'scrappy' has been running on FICS on a P5/133 computer, >>and this guy has played it a bunch of games. With only (maybe) one draw >>every 20 games or so... Not that 'crafty' (scrappy handle on FICS) is all >>that good, but it does have some distinct ideas/plans for preventing a blocked >>position which often leads to a draw. The only problem I haven't tackled yet >>is to recognize book lines that block the position too much (particularly >>things like white pawns at b3/c4/d5/e4 with black pawns at b4/c5/d6/e5. By >>then the options to prevent blocking things are almost gone, before my eval code >>has a chance to kick in... >> >>I'm not sure that 'not blocking' the position is a good idea, if the opponent is >>a computer, because that tends to avoid drawish positions. If your opponent is >>then tactically better than you, you might regret keeping it open (for him)... >> >>I will probably put scrappy back on the quad pentium-pro/200 on Monday. I was >>in the process of upgrading to redhat 6.1 but ran into a video glitch (the video >>card went south for the winter) and by the time the replacement arrived, I was >>in the hospital. > > >Nothing serious I hope. > > > > Christophe nope... relatively minor surgery to fix something that has been causing me problems for several years. I'm home, on my notebook, feeling OK... Now that I am 'fixed' it is time to get crafty 'fixed' again... It has some quirks I have recently introduced that need maybe major surgery. :) > > > >> It has done worse against computers (several PIII/500's and >>600's are beating up on it pretty good) but against humans there hasn't been a >>noticable decrease in wins... >> >>But if you don't specifically prevent blocking the position, it will happen. >>Kingway (IM Kim Commons) is a master at blocking things up. I'm nearly a master >>at preventing it. :) >> >>Bob >> >> >> >> >> >>>[Event "ICS Game"] >>>[Site "ICS"] >>>[Date "1999.12.02"] >>>[White "shutka"] >>>[Black "chesspartner"] >>> >>>1. d4 d5 2. c3 Nf6 3. Bf4 c6 4. Qb3 e6 5. Bg3 Ne4 6. Nf3 c5 7. e3 Nxg3 8. hxg3 >>>c4 9. Qc2 h6 10. Nbd2 a6 11. Be2 Nc6 12. O-O Bd6 13. Rfe1 b5 14. Bf1 O-O 15. e4 >>>Re8 16. e5 Bc7 17. Be2 Bb6 18. Nf1 b4 19. a3 b3 20. Qd2 Kf8 21. Ne3 Ra7 22. Nh2 >>>Qg5 23. Nf3 Qg6 24. Rac1 Rb7 25. Nh2 Rc7 26. Nf3 Rd7 27. Nh2 a5 28. Nf3 Bb7 >>>29. Nh2 Ra8 30. Nf3 a4 31. Nh2 Qg5 32. Nf3 Qd8 33. Nh2 Ba6 34. Nf3 Bc8 35. Nh2 >>>Kg8 36. Nf3 Ba5 37. Nh2 Bb7 38. Nf3 Ba6 39. Nh2 Bb6 40. Nf3 Qc8 41. Nh2 Bb5 >>>42. Nf3 Bd8 43. Nh2 Bg5 44. f4 Bd8 45. Nf3 Rb7 46. Rf1 Qd7 47. Rf2 Kf8 48. Rcf1 >>>Rbb8 49. Re1 Ke8 50. Ref1 Ba6 51. Re1 Bc8 52. Ref1 Kf8 53. Re1 Kg8 54. Ref1 >>>Kh7 55. Re1 Ne7 56. Ref1 Qe8 57. Re1 Kg8 58. Ref1 Kf8 59. Re1 Qd7 60. Ref1 >>>Kg8 61. Re1 Qb5 62. Ref1 Nc6 63. Re1 Kh7 64. Ref1 Rb7 65. Re1 Bc7 66. Ref1 >>>Rbb8 67. Re1 Ne7 68. Ref1 Qc6 69. Re1 Qe8 70. Ref1 Bd7 71. Re1 Bc6 72. Ref1 >>>Qc8 73. Re1 Qd8 74. Ref1 Kg8 75. Re1 Ra7 76. Ref1 Ba5 77. Re1 Bb6 78. Ref1 >>>Kh7 79. Re1 Bb5 80. Ref1 Rbb7 81. g4 Ng6 82. g3 Ra8 83. Re1 f6 84. Ref1 fxe5 >>>85. Nxe5 Nxe5 86. fxe5 Qg5 87. Kg2 Kg8 88. Bf3 Rf8 89. Be2 Rbf7 90. Rxf7 Rxf7 >>>91. Rxf7 Kxf7 92. Bf3 Kg8 >>> >>>Thanks.
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.