Author: enrico carrisco
Date: 21:57:13 11/20/05
Go up one level in this thread
On November 21, 2005 at 00:07:38, A. Steen wrote: >On November 20, 2005 at 23:23:14, enrico carrisco wrote: > >>On November 20, 2005 at 23:03:44, A. Steen wrote: >> >>>On November 20, 2005 at 22:42:02, enrico carrisco wrote: > > > ::: snipped ten examples of engines analysing a KRK position; all found the >best move (Ke3) immediately in a tiny fraction of a second and gave it a high >evaluation ::: > > >>>>Please -- will the die-hard "EGTB files are a waste" believers please stand-up. >>>> >>>>Regards, >>>> >>>>Enrico. >>> >>> >>>I think EGTBs are useful, but I think your examples illustrate the opposite of >>>what you are trying to prove. :) >>> >> >>Or perhaps you miss my point and need a clearer path of explanation. > >We'll see what I missed. :) Or what I missed... And why it is clear to me that I should have put forth a more solid (and clear) foundation to begin with. Then, the obvious point that the position is "won" would not arise. http://www.talkchess.com/forums/1/message.html?463265 > >>>Good chess engines are not optimised mate-finders, but optimised win-finders, >>>and there is a big difference! >> >>So, after viewing the example below, your definition of win-finder would be >>taking the scenic route? Your statement is nonsense as finding mate is >winning. > >My statement is not "nonsense". :) > >You are confusing at least four potentially distinct things: >*a) winning by the fastest method available, and knowing it is fastest and a >100% sure win (i.e. a seen mate) >*b) winning by the fastest method available but not knowing it is fastest or a >100% sure win >*c) winning by a more "scenic route" and knowing it is a 100% sure win (a seen >mate) >*d) winning by a more "scenic route" and not knowing it is a 100% sure win. > >A tablebase result would be of type (a). > >All your chosen engine analysis examples were of type (b), with the move found >(and potentially made, if the TC was demanding enough) as fast as a tablebase >lookup would be. Whether directly (root position) or while carrying out an >evaluation/search. > >So in the single position example given by you, the EGTBs have no advantage. > >Had you chosen examples of types (c) or (d), you may likely have had a point. > >But you did not. > >Whether the program knows or not it is a fastest win is usually irrelevant as >long as it makes the perfect move fast enough. As Rhett Butler put it: "Frankly >my dear, I don't give a damn." In a manner of speaking. So, in your words, it is unimportant if 21 extra moves are made as long as they are made "fast enough." Unfortunately, I think many would "give a damn", especially if trying to analyze a position for greater insight. (Especially a self-proclaimed "almost patzer.") Or is it genius? (See: http://www.talkchess.com/forums/1/message.html?463281) Here's another example -- perhaps you will find it a worthy one: With EGTBs it finds an instant mate in 14 (Rg6.) Without, however: [D]8/1k6/8/8/8/8/6R1/7K w - - 0 1 Analysis by Fruit 2.2.1: 1.Kg1 +- (5.00) Depth: 1/1 00:00:00 1.Kg1 Kc6 +- (5.03) Depth: 2/4 00:00:00 1.Kg1 Kc6 2.Kf1 +- (4.95) Depth: 3/6 00:00:00 1.Rb2+ Kc6 2.Kg2 Kd5 3.Rb5+ Kd4 +- (5.09) Depth: 3/8 00:00:00 1.Rb2+ Kc6 2.Kg2 Kd5 3.Rb5+ Kd4 +- (5.09) Depth: 4/9 00:00:00 1.Kg1 Kc6 2.Kf2 Kd6 3.Rg6+ Kd5 +- (5.18) Depth: 4/9 00:00:00 1.Rg6 Kc7 2.Kg2 Kd7 +- (5.21) Depth: 4/9 00:00:00 1.Rg6 Kc7 2.Kg2 Kd7 3.Kf3 +- (5.25) Depth: 5/9 00:00:00 1.Rb2+ Kc6 2.Kg2 Kd6 3.Kf3 Ke5 4.Re2+ Kd5 +- (5.28) Depth: 5/10 00:00:00 1.Rb2+ Kc6 2.Kg2 Kd6 3.Kf3 Ke5 4.Re2+ Kd5 +- (5.28) Depth: 6/12 00:00:00 21kN 1.Rg6 Kc7 2.Kg2 Kd7 3.Kf3 Ke7 +- (5.47) Depth: 6/12 00:00:00 22kN 1.Rg6 Kc7 2.Kh2 Kb7 3.Kg3 Kc8 4.Rg7 +- (5.73) Depth: 7/12 00:00:00 52kN 1.Rg6 Kc7 2.Kh2 Kb7 3.Kg3 Kc8 4.Rg7 Kd8 +- (5.86) Depth: 8/14 00:00:01 156kN 1.Rg6 Kc7 2.Kh2 Kb7 3.Kg3 Kc8 4.Rg7 Kd8 5.Kf4 +- (5.95) Depth: 9/16 00:00:01 472kN 1.Rg6 Kc8 2.Rg7 Kd8 3.Kg2 Ke8 4.Kf2 Kd8 5.Ke3 Ke8 +- (6.16) Depth: 10/18 00:00:01 1323kN 1.Rg6 Kc8 2.Rg7 Kd8 3.Kg2 Ke8 4.Rb7 Kf8 5.Kg3 Kg8 6.Kf4 +- (6.14) Depth: 11/18 00:00:01 1560kN 1.Rg6 Kc7 2.Kh2 Kb7 3.Rh6 Ka7 4.Kg2 Kb8 5.Kf3 Kc8 6.Rh7 Kd8 +- (6.06) Depth: 12/20 00:00:04 9878kN 1.Rg6 Kc7 2.Kg2 Kd7 3.Kf3 Ke7 4.Ke4 Kf7 5.Rd6 Kg7 6.Kf5 Kf7 7.Rd7+ Ke8 +- (6.16) Depth: 13/22 00:00:14 34907kN 1.Rg6 Kc7 2.Kg2 Kd7 3.Kf3 Ke7 4.Ra6 Kd7 5.Ke4 Kc7 6.Kd3 Kd8 7.Ra7 Ke8 +- (6.16) Depth: 14/24 00:00:23 55353kN 1.Rg6 Kc7 2.Kg2 Kd7 3.Kf3 Ke7 4.Kf4 Kf7 5.Rc6 Ke7 6.Ke5 Kd7 7.Rc4 Ke7 8.Rc7+ Kd8 +- (6.25) Depth: 15/24 00:00:46 109994kN 1.Rg6 Kc7 2.Kg2 Kd7 3.Kf3 Ke7 4.Kf4 Kf7 5.Rd6 Kg7 6.Rb6 Kh7 7.Ra6 Kg8 8.Ra7 Kf8 +- (6.26) Depth: 16/26 00:01:54 270223kN 1.Kg1 Ka6 2.Kf2 Ka5 3.Rg5+ Kb6 4.Ke3 Kc7 5.Rb5 Kd8 6.Rd5+ Ke8 7.Kf4 Kf8 8.Re5 Kg8 9.Re7 Kf8 +- (6.27) Depth: 16/26 00:02:41 389657kN 1.Rg4 Kb6 2.Kg2 Kc6 3.Kf3 Kd7 4.Rd4+ Ke6 5.Kf4 Ke7 6.Kf5 Kf7 7.Rd7+ Ke8 8.Ke6 Kf8 9.Rh7 Kg8 +- (6.32) Depth: 16/26 00:03:11 463143kN 1.Rg4 Kb6 2.Kg2 Kc6 3.Kf3 Kd7 4.Rc4 Ke6 5.Kf4 Kd7 6.Ke5 Ke7 7.Rc7+ Kd8 8.Rg7 Kc8 9.Kd6 Kb8 +- (6.32) Depth: 17/28 00:03:49 552376kN 1.Rb2+ Kc6 2.Kg2 Kd6 3.Kf3 Kd7 4.Kf4 Kd6 5.Rd2+ Ke6 6.Rd3 Ke7 7.Ke5 Kf7 8.Rf3+ Kg6 9.Rf6+ Kg7 10.Ke6 Kh8 11.Rf7 +- (6.34) Depth: 17/28 00:06:11 900639kN 1.Rb2+ Kc6 2.Kg2 Kd6 3.Kf3 Kd7 4.Ke4 Kc6 5.Ke5 Kd7 6.Rc2 Ke7 7.Rc7+ Kd8 8.Rg7 Kc8 9.Kd6 Kb8 10.Rc7 Ka8 +- (6.55) Depth: 18/29 00:09:05 1339108kN -elc. >> By the way -- your profile is missing. > >Excellent, and most highly relevant to the discussion. > >>>So whether the programs found mate instantly, let alone the quickest mate, or >>>not is irrelevant. >>> >>>In all your examples, the programs found the fastest winning move (Ke3 - and I >>>don't need to check that, do I - surely everything else must be at least 1 move >>>slower?) in much less than one second. Probably in just a few thousandths of a >>>second. And all gave it a high evaluation. Only one (J9) ever wavered from >>>that choice Ke3. >>> >>>So, even if EGTBs were being used in the search itself, the presence of EGTBs >>>would not have helped at all in this example. Chess Challenger 7 could not >>>handle KRK, but we have moved on since then. >>> >>>Simply put, you need a more complex example. :) >> >>I wasn't trying to hang Fruit with my example, so I didn't hammer my point clear >>enough, perhaps. Yes it found the next suceeding move but not knowing a clear >>path to mate in very simplistic ending can end with a lot of wasted time -- even a flag. > >Some endings - many endings - only look simplistic. > >>Look at moves 86. on in the following exmaple. Fruit 2.2.1 was black. > >OK. But a bit earlier, at 58., there is what I think may be a blunder by white. > > >> >>1. Nf3 {0:05:03} Nf6 {0:05:03} 2. c4 {0:05:06} c5 {0:05:06} 3. g3 {0:05:08} >>d5 {0:05:09} 4. cxd5 {0:05:11} Nxd5 {0:05:11} 5. Bg2 {0:05:14} Nc6 {0:05:14} >>6. Nc3 {0:05:17} Nc7 {0:05:17} 7. Qa4 {0:05:20} Bd7 {0:05:20} 8. Qe4 >>{0:05:23} g6 {0:05:23} 9. Ne5 {0:05:26} Bg7 {0:05:26} 10. Nxd7 {0:05:29} >>Qxd7 {0:05:29} 11. O-O {0:05:32} O-O {0:05:32} 12. a3 {0:05:35} Rac8 >>{0:05:34} 13. b4 {0:05:37} cxb4 {0:05:23} 14. axb4 {0:05:25} Nb5 {0:05:25} >>15. Bb2 {0:05:07} a6 {0:05:28} 16. Rfd1 {0:04:41} Nd6 {0:05:31} 17. Qd5 >>{0:04:25} Rfd8 {0:05:34} 18. Qa2 {0:04:14} e6 {0:05:28} 19. Rdc1 {0:04:16} >>Nf5 {0:05:19} 20. Bxc6 {0:04:16} Rxc6 {0:05:00} 21. d3 {0:04:00} Nd4 >>{0:04:34} 22. Kg2 {0:03:54} e5 {0:04:19} 23. Rab1 {0:03:30} h5 {0:04:22} 24. >>Ne4 {0:03:25} Nxe2 {0:04:21} 25. Rxc6 {0:03:18} Qxc6 {0:04:17} 26. Qc4 >>{0:03:14} Qd5 {0:04:16} 27. Qxd5 {0:03:08} Rxd5 {0:04:19} 28. Kf3 {0:03:04} >>Nd4+ {0:04:12} 29. Bxd4 {0:03:06} Rxd4 {0:04:01} 30. Ke3 {0:03:09} b6 >>{0:03:46} 31. Nd2 {0:03:01} Bh6+ {0:03:45} 32. Ke2 {0:03:00} Bxd2 {0:03:43} >>33. Kxd2 {0:02:55} Kg7 {0:03:43} 34. Kc3 {0:02:46} Rd6 {0:03:42} 35. Re1 >>{0:02:42} Kf6 {0:03:30} 36. Re4 {0:02:45} Kf5 {0:03:22} 37. f4 {0:02:39} >>Rc6+ {0:03:23} 38. Kd2 {0:02:30} exf4 {0:03:18} 39. Rxf4+ {0:02:32} Ke6 >>{0:03:14} 40. Re4+ {0:02:32} Kd7 {0:03:07} 41. Re1 {0:02:28} a5 {0:02:59} >>42. Ra1 {0:02:28} axb4 {0:02:52} 43. Rb1 {0:02:31} Rf6 {0:02:46} 44. Ke3 >>{0:02:30} Re6+ {0:02:39} 45. Kf4 {0:02:33} Re2 {0:02:32} 46. Rxb4 {0:02:28} >>Rf2+ {0:02:26} 47. Ke4 {0:02:19} Rxh2 {0:02:29} 48. Rxb6 {0:02:13} Rh3 >>{0:02:32} 49. Kf4 {0:02:07} h4 {0:02:31} 50. gxh4 {0:02:05} Rxd3 {0:02:26} >>51. Kg5 {0:02:06} Rf3 {0:02:14} 52. Rb7+ {0:02:02} Ke6 {0:02:17} 53. Rb6+ >>{0:01:54} Ke7 {0:02:19} 54. Rb7+ {0:01:51} Kf8 {0:02:10} 55. Rb5 {0:01:54} >>Rh3 {0:02:05} 56. Kf6 {0:01:56} Kg8 {0:02:03} 57. Rb8+ {0:01:57} Kh7 >>{0:02:05} 58. Kxf7 {0:01:54} Rxh4 {0:02:03} 59. Kf6 {0:01:57} Rh5 {0:02:05} >>60. Rb1 {0:01:52} Rf5+ {0:02:08} 61. Ke6 {0:01:47} Kh6 {0:02:10} 62. Rb2 >>{0:01:42} Rf4 {0:02:13} 63. Rh2+ {0:01:38} Kg5 {0:02:15} 64. Ke5 {0:01:28} >>Rh4 {0:02:18} 65. Rf2 {0:01:23} Kh5 {0:02:21} 66. Rf8 {0:01:22} g5 {0:02:23} >>67. Rh8+ {0:01:17} Kg4 {0:02:26} 68. Rb8 {0:01:15} Rh1 {0:02:29} 69. Kf6 >>{0:01:13} Rh6+ {0:02:31} 70. Ke5 {0:01:11} Rc6 {0:02:34} 71. Ke4 {0:01:04} >>Re6+ {0:02:36} 72. Kd4 {0:01:03} Re1 {0:02:39} 73. Rh8 {0:00:58} Kg3 >>{0:02:41} 74. Kd3 {0:00:55} g4 {0:02:44} 75. Rh6 {0:00:50} Kg2 {0:02:46} 76. >>Rf6 {0:00:44} g3 {0:02:48} 77. Rf7 {0:00:41} Re5 {0:02:51} 78. Rh7 {0:00:41} >>Kf3 {0:02:54} 79. Rf7+ {0:00:41} Kg4 {0:02:56} 80. Rf1 {0:00:42} g2 >>{0:02:58} 81. Rb1 {0:00:19} Kf3 {0:03:01} 82. Kd4 {0:00:17} Rg5 {0:03:03} >>83. Rg1 {0:00:04} Kf2 {0:03:06} 84. Rd1 {0:00:05} g1=Q {0:03:08} 85. Rxg1 >>{0:00:08} Rxg1 {0:03:02} 86. Kd5 {0:00:07} Re1 {0:02:48} 87. Kd4 {0:00:05} >>Rd1+ {0:02:26} 88. Ke4 {0:00:04} Rd8 {0:02:19} 89. Ke5 {0:00:06} Kf3 >>{0:02:00} 90. Ke6 {0:00:09} Ke4 {0:01:44} 91. Kf6 {0:00:07} Kd5 {0:01:22} >>92. Ke7 {0:00:05} Rc8 {0:01:11} 93. Kf6 {0:00:07} Rc3 {0:01:03} 94. Kf5 >>{0:00:07} Rh3 {0:00:57} 95. Kf4 {0:00:07} Rh4+ {0:00:56} 96. Kf5 {0:00:09} >>Rh5+ {0:00:52} 97. Kf4 {0:00:07} Re5 {0:00:41} 98. Kf3 {0:00:05} Rf5+ >>{0:00:36} 99. Ke3 {0:00:04} Rf7 {0:00:32} 100. Kd2 {0:00:04} Kc5 {0:00:24} >>101. Ke3 {0:00:05} Rf1 {0:00:19} 102. Ke4 {0:00:04} Kd6 {0:00:11} 103. Kd3 >>{0:00:04} Rf3+ {0:00:05} 104. Ke4 {0:00:04} Rf8 {0:00:04} 105. Kd4 {0:00:04} >>Rf4+ {0:00:07} 106. Kd3 {0:00:05} Kc5 {0:00:04} 107. Ke3 {0:00:06} Ra4 >>{0:00:04} 108. Kd3 {0:00:06} Rd4+ {0:00:07} 109. Ke3 {0:00:06} Rb4 {0:00:04} >>110. Kd3 {0:00:06} Rh4 {0:00:04} 111. Ke3 {0:00:06} Kc4 {0:00:05} 112. Kf3 >>{0:00:08} Kd3 {0:00:04} 113. Kg3 {0:00:06} Rb4 {0:00:07} 114. Kf3 {0:00:05} >>Kd2 {0:00:10} 115. Kf2 {0:00:05} Rb3 {0:00:12} 116. Kf1 {0:00:07} Ke3 >>{0:00:12} 117. Kg2 {0:00:10} Ke2 {0:00:13} 118. Kg1 {0:00:13} Kf3 {0:00:14} >>119. Kh2 {0:00:15} Rb1 {0:00:14} 120. Kh3 {0:00:18} Rh1# {0:00:16} >>{White checkmated} >>0-1 > >Well, EGTBs hook on to the win once white allows the pawn exchange at 58., >giving m/44. So at 86, with optimal play for both, we should be at about m/16. >In fact we are at m/13 as white has played worse. 86 + 13 = 99. So there was >an extension by 120-99 = 21 moves, caused by stupid play by black starting with >92 and then getting worse. > >Yes, this is a better example from you. :) > >It illustrates that repeated "scenic routes" can literally result in running >around in pointless rectangles (like the BR does). > >Your single position example did not. > >Aside: EG play by both Fruity and Fritzy9 is sometimes suspect. It comes with >the territory of making the progs smarter, maybe. > >I am on the side of EGTBs. > >Best, > >A.S.
This page took 0.01 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.