Author: Uri Blass
Date: 05:11:48 02/09/03
Go up one level in this thread
On February 08, 2003 at 13:47:29, Drexel,Michael wrote: >[D] 3k3r/3p3p/pr1Pp3/2p1nppB/2P5/1P2P3/P2R1KPP/5R2 w - - 0 26 > >black has advantages and white has to play precisely in order to survive. >the best move is probably 26.g4. >Comet and several other freeware engines dont understand the position and >favour white: > >Analysis by Comet B55: > >1. ² (0.54): 26.e4 Tf8 27.Ke2 f4 28.h3 a5 >2. ² (0.54): 26.Ke2 a5 27.e4 f4 28.g3 fxg3 >3. ² (0.54): 26.Tdd1 Tf8 27.Ke2 g4 28.Kf2 Tg8 29.Kg1 Tf8 30.Td2 >4. ² (0.54): 26.Kg3 a5 27.Tc1 Tf8 28.Kh3 >5. ² (0.54): 26.h3 a5 27.e4 Tf8 28.Kg1 a4 29.Lf3 >6. ² (0.54): 26.Tc1 Tf8 27.h3 > >3:03 > >Comet B55 - Fritz 8 >3k3r/3p3p/pr1Pp3/2p1nppB/2P5/1P2P3/P2R1KPP/5R2 w - - 0 1 > >Analysis by Fritz 8: > >1. ³ (-0.28): 26.g4 fxg4 27.Kg3 a5 28.Lxg4 a4 29.Le2 axb3 30.axb3 Kc8 31.Ld1 >2. ³ (-0.41): 26.h4 h6 27.g3 Tf8 28.Kg2 Tg8 29.hxg5 hxg5 30.g4 Th8 31.Kg3 a5 >3. ³ (-0.44): 26.h3 a5 27.Le2 Tf8 28.Tfd1 Tb4 29.g3 h5 30.Lxh5 a4 >4. ³ (-0.44): 26.Kg3 a5 27.Le2 Tf8 28.h3 h6 29.Kf2 f4 >5. ³ (-0.44): 26.Tc1 a5 27.h3 Tf8 28.Le2 Tf6 29.g3 Sf7 30.Tcd1 e5 31.Td5 >6. ³ (-0.47): 26.Th1 Tf8 27.h4 h6 28.hxg5 hxg5 29.e4 f4 30.Lf3 Sf7 31.Th5 Tg8 > >3:06 > >After the mistake 26.e4? Comet was lost. the knight is much stronger of course. > >[Event "ATHLONXP, Blitz:5'+3"] >[Site "ATHLONXP"] >[Date "2003.02.08"] >[Round "1"] >[White "Comet B55"] >[Black "Fritz 8"] >[Result "0-1"] >[ECO "B49"] >[PlyCount "100"] >[EventDate "2003.??.??"] > >1. e4 c5 2. Nf3 e6 3. d4 cxd4 4. Nxd4 Nc6 5. Nc3 Qc7 6. Be3 a6 7. Be2 Nf6 8. >O-O Bb4 9. Na4 Be7 10. Nxc6 bxc6 11. Nb6 Rb8 12. Nxc8 Qxc8 13. e5 Nd5 14. Bc1 >Bc5 15. c4 Ne7 16. b3 Ng6 17. Be3 Bxe3 18. fxe3 Qc7 19. Qd6 Qxd6 20. exd6 c5 >21. Rae1 Rb6 22. Rd1 Ne5 23. Kf2 f5 24. Rd2 g5 25. Bh5+ Kd8 26. e4 f4 27. g3 >Rf8 28. Kg2 a5 29. h4 h6 30. Be2 Ra6 31. Rf2 Kc8 32. hxg5 hxg5 33. Rf1 Kb7 34. >Rf2 a4 35. Rf1 Kc6 36. Rf2 axb3 37. axb3 Ra1 38. Bd1 Rc1 39. gxf4 gxf4 40. Rf1 >Rc3 41. Rh1 Re3 42. Kf1 Rxe4 43. Rh7 f3 44. Rh5 Ng4 45. Bc2 Ref4 46. Rh3 Ra8 >47. Rh1 Ra1+ 48. Rd1 Ra2 49. Rd2 Rxc2 50. Rxc2 Ne3+ 0-1 I think that Comet has a big problem with it's search rules(see analysis at the bottom of this post). Unfortunately after 2 hours of search I decided to stop it. It's pv at depth 15 does not make and I do not understand 2...Nxc4 and 3...Rxd6 at that depth. If it is a result of MTD then I think that it may be better if Comet stop to use MTD because if it starts to search first the pv moves then it is going to have bad order of moves in the next iteration. I also do not think that all the 0.01 windows are a good idea because I suspect that the result is that Comet is slower in getting another ply deeper at least in the position that was analyzed. Let look at the scores for e4. There is a clear patern of going down and I think that it may be better to teach Comet to learn from the patterns that it see and avoid windows of 0.01 at least in cases that experience proves that the real difference is usually more than 0.01. 9 ->0.42 10->0.41->0.39 11->0.38->0.32 12->0.32->0.32(I wonder why 2 times) 13->0.31->0.25 14->0.24->0.21 15->0.20->0.20 16->0.19->? New game 3k3r/3p3p/pr1Pp3/2p1nppB/2P5/1P2P3/P2R1KPP/5R2 w - - 0 1 Analysis by Comet B55: 1.Bg6 hxg6 -+ (-3.77) Depth: 1/3 00:00:00 1.Rdd1 Rf8 2.Ke2 ² (0.59) Depth: 3/8 00:00:00 1.Rdd1 Rf8 2.Ke2 ² (0.59) Depth: 3/8 00:00:00 1.Rdd1 Rf8 2.Ke2 ² (0.59) Depth: 3/8 00:00:00 1.Rdd1 Rf8 2.Ke2 ² (0.59) Depth: 3/8 00:00:00 1.Rdd1 Rf8 2.Ke2 ² (0.59) Depth: 3/8 00:00:00 1.Rdd1 Rf8 2.Ke2 ² (0.59) Depth: 3/8 00:00:00 1.Rdd1 Rf8 2.Ke2 ² (0.59) Depth: 3/8 00:00:00 1.Rdd1 Rf8 2.Ke2 ² (0.59) Depth: 3/8 00:00:00 1.g3 g4 2.Rdd1 ² (0.60) Depth: 3/8 00:00:00 1.g3 g4 2.Rdd1 Rf8 ² (0.47) Depth: 4/8 00:00:00 1.Rdd1 Rf8 2.Bf3 Nxf3 3.Kxf3 ² (0.59) Depth: 4/8 00:00:00 1.e4 fxe4 2.Ke3 Re8 ² (0.60) Depth: 4/10 00:00:00 1.e4 Rg8 2.Ke3 f4+ 3.Ke2 ² (0.62) Depth: 4/10 00:00:00 1.e4 h6 2.exf5 Rf8 3.Rfd1 exf5 ² (0.61) Depth: 5/12 00:00:00 12kN 1.e4 h6 2.exf5 Rf8 3.Ke3 exf5 ² (0.53) Depth: 5/12 00:00:00 22kN 1.Rdd1 Rf8 2.Kg3 Rf6 3.Bf3 ² (0.54) Depth: 5/12 00:00:00 25kN 1.Rdd1 Rf8 2.g4 fxg4+ 3.Kg1 Rxf1+ 4.Kxf1 ² (0.53) Depth: 6/17 00:00:00 39kN 1.Rdd1 Rf8 2.Kg3 Kc8 3.e4 f4+ 4.Kf2 ² (0.51) Depth: 6/17 00:00:00 78kN 1.Rdd1 Rf8 2.Kg3 Kc8 3.Rd2 ² (0.50) Depth: 7/17 00:00:01 121kN 1.Ke2 g4 2.e4 Rf8 3.Ke3 ² (0.51) Depth: 7/19 00:00:02 244kN 1.Ke2 g4 2.e4 Rf8 3.Ke3 Rf6 4.exf5 Rxf5 5.Rxf5 exf5 ² (0.51) Depth: 7/19 00:00:02 248kN 1.Ke2 Rf8 2.h3 Rf6 3.Rfd1 ² (0.50) Depth: 8/19 00:00:02 323kN 1.Ke2 Rf8 2.e4 f4 3.h3 h6 4.Rdd1 Kc8 ² (0.41) Depth: 8/21 00:00:04 635kN 1.Ke2 a5 2.e4 f4 3.Rxf4 gxf4 4.Be8 ² (0.40) Depth: 9/21 00:00:08 1374kN 1.e4 Rf8 2.Ke2 f4 ² (0.41) Depth: 9/22 00:00:10 1774kN 1.e4 Rf8 2.Ke2 f4 3.h3 h6 4.Kd1 Kc8 5.Rff2 ² (0.42) Depth: 9/22 00:00:10 1895kN 1.e4 Rf8 2.Ke2 f4 3.h3 h6 4.g3 ² (0.41) Depth: 10/26 00:00:16 2905kN 1.e4 Rf8 2.Kg1 f4 3.h3 h6 4.Rdd1 Kc8 5.Kh2 Rb4 ² (0.39) Depth: 10/26 00:00:29 5362kN 1.e4 Rf8 2.Kg1 f4 3.h3 a5 4.Rxf4 gxf4 5.Bf3 ² (0.38) Depth: 11/26 00:00:41 7751kN 1.e4 Rf8 2.Kg1 f4 3.h3 a5 4.Be2 Rb4 5.Rb2 h6 6.Kh2 ² (0.32) Depth: 11/26 00:01:07 13171kN 1.e4 Rf8 2.Ke2 g4 3.Ke3 Nc6 ² (0.32) Depth: 12/28 00:02:28 31028kN 1.e4 Rf8 2.Ke2 f4 3.h3 a5 4.Rc1 a4 5.Rdd1 axb3 6.axb3 ² (0.32) Depth: 12/28 00:02:32 31981kN 1.e4 Rf8 2.Ke2 f4 3.h3 a5 4.Rc1 Rb4 5.Kd1 a4 6.Bf7 ² (0.31) Depth: 13/28 00:03:29 43917kN 1.e4 Rf8 2.Kg1 f4 3.h3 a5 4.Be2 Rb4 5.Rdd1 a4 6.Rb1 h6 7.bxa4 Rxa4 8.Rb8# = (0.25) Depth: 13/30 00:08:16 110392kN 1.e4 f4 2.Kg1 a5 3.Be2 Rb4 4.g3 a4 5.gxf4 gxf4 6.Rxf4 axb3 7.Rg4 = (0.24) Depth: 14/33 00:18:47 257666kN 1.e4 f4 2.h3 a5 3.Rc1 Rb4 4.Rc3 h6 5.a3 Rb6 6.Ke2 Kc8 7.g3 fxg3 8.Rxg3 = (0.21) Depth: 14/34 00:39:51 540586kN 1.e4 f4 2.h3 a5 3.Rc1 Kc8 4.Kg1 Rb4 5.Re1 h6 6.Be2 a4 7.Rf1 = (0.20) Depth: 15/34 01:03:46 885784kN 1.e4 f4 2.h3 Nxc4 3.bxc4 Rxd6 4.Rxd6 h6 = (0.20) Depth: 15/37 01:45:34 1479984kN 1.e4 f4 2.h3 a5 3.Be2 Rf8 4.Kg1 h6 5.Rfd1 Rb4 6.Rb2 f3 7.gxf3 Nxf3+ 8.Kh1 = (0.19) Depth: 16/37 01:57:29 1655963kN (Blass, Tel-Aviv 09.02.2003) 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.