Author: Uri Blass
Date: 04:58:54 05/13/02
Go up one level in this thread
On May 13, 2002 at 07:34:57, Ulrich Tuerke wrote: >On May 13, 2002 at 06:39:57, Uri Blass wrote: > >>On May 13, 2002 at 06:31:08, Jorge Pichard wrote: >> >>>On May 12, 2002 at 22:21:39, Christophe Theron wrote: >>> >>>>On May 12, 2002 at 14:49:26, Tony Hedlund wrote: >>>> >>>>>[Event "Tony Hedlund"] >>>>>[Site "SSDF"] >>>>>[Date "2002.05.12"] >>>>>[Round "4"] >>>>>[White "Chess Tiger 14.9 Palm"] >>>>>[Black "Comet A90 P200"] >>>>>[BlackElo "2253"] >>>>>[Result "0-1"] >>>>>[Opening "A30 English: Symmetrical Variation"] >>>>> >>>>>1. c4 c5 2. d3 Nc6 3. Nf3 d6 4. e4 h6 5. h3 Nf6 6. Nc3 e5 7. Nd5 Bd7 >>>>>8. Bd2 Be7 9. Nxe7 Qxe7 10. a3 O-O 11. Be2 Rfb8 12. Bc3 b5 13. cxb5 Rxb5 >>>>>14. O-O a6 15. Qd2 Rbb8 16. Bd1 a5 17. Re1 a4 18. Rc1 Rb7 19. Bc2 Qe6 >>>>>20. Bd1 Nd4 21. Kf1 Rab8 22. Rb1 Qa2 23. Rc1 Rb6 24. Qe3 Bc6 25. Nd2 Bb5 >>>>>26. Be2 Qe6 27. Nc4 R6b7 28. Red1 Nb3 29. Rc2 Ba6 30. Kg1 Nd4 31. Rcc1 d5 >>>>>32. Bxd4 cxd4 33. exd5 Nxd5 34. Qg3 Bxc4 35. Rxc4 Nf4 36. Bg4 Qa6 >>>>>37. Re1 Rb3 38. Bf5 Qa5 39. Kf1 Rxb2 40. Rec1 g6 41. Rd1 Qd2 42. Qf3 Rb1 >>>>>43. Bg4 e4 44. dxe4 R8b2 45. Rc8+ Kg7 46. Kg1 Ne2+ 47. Kh2 Rxd1 48. g3 Nc3 >>>>>49. Kg2 Qe1 50. Qxf7+ Kxf7 51. Rc7+ Kf8 52. Rc8+ Ke7 53. Rc7+ Kd6 >>>>>54. Rd7+ Kc6 55. Rd6+ Kxd6 56. e5+ Kxe5 57. Bxd1 Rxf2# 0-1 >>>>> >>>>>Tiger was out of book after 1.c4 c5! Please Chris, a BIGGER book! >>>>>Tal was out of book after 2.d3. >>>>> >>>>>Tony >>>> >>>> >>>> >>>>The small book that Chess Tiger for Palm is using is an improved version of the >>>>book used by Chess Tiger 11.8 when the SSDF has tested it! >>>> >>>>With this book, Tiger 11.8 got the first place amongst programs tested on P90. >>>> >>>>But yes, it would help if this book was bigger. >>>> >>>> >>>> >>>> Christophe >>> >>>You finally realized it. I have been saying this for over three months. >>> >>>Jorge. >> >>Christophe said a long time ago that adding a big book for palm tiger is not an >>easy task because of some design decision. >> >>I also do not like a big book because it may prevent me to see problems of the >>engine in the opening. >> >>I believe that a good engine should find good moves at the opening stage with no >>book. > >Nevertheless, you need a minimum of diversity in the opening book. >Imagine Palm-CT against Comet A.90 would next time start again with >1.c4 c5 >Tiger would then have "best chances" to reproduce this loss. It is possible to prevent this problem without book by small randomness in the evaluation or some learning to dislike moves that were played in the past by remembering their score and playing them again only if you do not find a move that is at least 0.1 pawns weaker. I guess that after 1.c4 c5 tiger can find a different move than 2.d3 if it is using that learning. I do not think that 2.d3 is more than 0.1 pawns better than all the other moves. Note that I did not implement it in my engine but I prefer implementing something like this and not to implement a big book at least in the near future. Uri
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.