Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: How much does tablebases affect the playing strength of Crafty

Author: Jerry Adams

Date: 16:24:32 05/17/00

Go up one level in this thread


On May 17, 2000 at 18:11:08, Robert Hyatt wrote:

>On May 17, 2000 at 15:53:21, Jerry Adams wrote:
>
>>On May 17, 2000 at 15:34:05, Robert Hyatt wrote:
>>
>>>On May 17, 2000 at 15:10:51, Jerry Adams wrote:
>>>
>>>>Hi
>>>>
>>>> My Crafty has been losing won positions because of it's horrible endgame. At
>>>>present I have absolutely no tablebases. I was under the impression that
>>>>tablebases would only account for a few additional rating points. Iv'e seen
>>>>crafty actually draw when it had a rook and King against a King!!! Question: How
>>>>can I build tablesbases without using alot of harddrive space? I only have a 4
>>>>gig hardrive. I would like crafty to be able to atleast win rudimentary
>>>>endgames. I am using winboard.
>>>
>>>
>>>If it draws with rook and king vs king, you have something broken.  It has
>>>specific code for such 'mating position'.  I have had it play hundreds of games
>>>with itself, one side using egtb's, the other not.  It has never failed to mate
>>>with KR vs K easily.  Not "optimally" usually, but quickly and easily.
>>>
>>>IMHO Crafty is not a "stupid program" in endings.  It handles itself pretty well
>>>most of the time.  Unless something is broken or mis-configured.  If you have a
>>>log.nnn file where it couldn't win KR vs K, that might tell what was happening.
>>
>>
>>
>> OKay here is the game played on ICC , Judge for yourself
>>
>>
>>
>>Event "ICS rated blitz match"]
>>[Site "chessclub.com"]
>>[Date "1998.12.05"]
>>[Round "?"]
>>[White "PostModernist"]
>>[Black "cmslim"]
>>[Result "1/2-1/2"]
>>[WhiteElo "2681"]
>>[BlackElo "2389"]
>>[ECO "C24"]
>>
>>1. e4 e5 2. Bc4 Nf6 3. d3 Nc6 4. Nf3 Be7 5. O-O O-O 6. Re1 d6 7. a4 h6
>>8. h3 Na5 9. Bd5 Nxd5 10. exd5 Bf6 11. Bd2 c6 12. Bxa5 Qxa5 13. dxc6 bxc6
>>14. Nc3 Rb8 15. Ne4 Be7 16. b3 Qc7 17. d4 d5 18. Nc3 e4 19. Ne5 Bb4 20.
>>Qd2 c5 21. Nxd5 Bxd2 22. Nxc7 Bc3 23. Nb5 Bxa1 24. Rxa1 cxd4 25. Nxa7 d3
>>26. Nac6 dxc2 27. Nxb8 Rd8 28. Nbd7 Bxd7 29. Rc1 Bxa4 30. Rxc2 Bxb3 31.
>>Rb2 Be6 32. Rb4 f6 33. Nc6 Rd6 34. Nd4 Bd7 35. Kh2 f5 36. h4 g5 37. h5
>>Kg7 38. Nb3 Be8 39. Nd4 Kf6 40. Rb8 Bf7 41. Nb5 Ra6 42. Nd4 f4 43. Rb5
>>Bxh5 44. Nf5 Ra2 45. Nxh6 Rxf2 46. Ng8+ Kf7 47. Rxg5 e3 48. Kg1 Bg6 49.
>>Nh6+ Kf6 50. Rc5 Re2 51. Ng4+ Kg7 52. Rc7+ Kg8 53. Rc8+ Kf7 54. Ne5+ Kf6
>>55. Nxg6 Kxg6 56. Kf1 Rd2 57. Re8 Kf5 58. Re7 Kg4 59. Rg7+ Kf5 60. Re7
>>Kg4 61. Rg7+ Kh5 62. Re7 Kh4 63. Rh7+ Kg5 64. Re7 Rc2 65. Re8 Ra2 66. Rb8
>>Rd2 67. Re8 Kg4 68. Rh8 Rb2 69. Rg8+ Kh5 70. Rh8+ Kg6 71. Rf8 Kg5 72. Rh8
>>Rf2+ 73. Kg1 Rd2 74. Rg8+ Kf5 75. Kf1 Rf2+ 76. Kg1 Ra2 77. Rf8+ Kg5 78.
>>Rg8+ Kf5 79. Rf8+ Ke4 80. Re8+ Kd3 81. Rd8+ Ke2 82. Rd4 f3 83. gxf3 Ra8
>>84. Rd5 Rg8+ 85. Kh2 Kf2 86. Kh3 e2 87. Re5 Rg3+ 88. Kh4 Rxf3 89. Rxe2+
>>Kxe2 90. Kg5 Kf2 91. Kg6 Kf1 92. Kg7 Re3 93. Kg8 Ke2 94. Kg7 Rg3+ 95. Kf7
>>Ke3 96. Kf8 Rf3+ 97. Ke8 Ke2 98. Ke7 Ke1 99. Kd6 Rg3 100. Kd5 Rd3+ 101.
>>Ke4 Ra3 102. Kd4 Ke2 103. Kd5 Ra7 104. Kd6 Kf3 105. Kc6 Kg3 106. Kd6 Ra1
>>107. Ke7 Kg4 108. Kd6 Kf3 109. Ke6 Kf4 110. Kd6 Ra7 111. Ke6 Rh7 112. Kd5
>>Rh1 113. Kd6 Kf5 114. Kd7 Kg6 115. Ke7 Rh7+ 116. Ke6 Rh1 117. Ke7 Rc1 118.
>>Ke6 Ra1 119. Ke5 Ra6 120. Ke4 Ra7 121. Ke5 Rh7 122. Ke6 Rh6 123. Ke5 Rh1
>>124. Ke4 Kf6 125. Kd4 Kf7 126. Ke4 Ke6 127. Kd4 Ke7 128. Kd5 Rh7 129. Kc5
>>Kf6 130. Kd5 Rh4 131. Kd6 Rh1 132. Kc6 Rc1+ 133. Kd6 Kf7 134. Kd7 Rb1 135.
>>Kd6 Ra1 136. Kc6 Ke7 137. Kc5 Ra2 138. Kb6 Kd8 139. Kc6 Rh2 1/2-1/2
>
>
>That looks like an old version (fairly old version) that had a bug in the
>evaluation code and never could call the EvaluateMate() function.  This was
>fixed several versions back, although I don't remember exactly when.
>
>Your game doesn't provide (1) which version and (2) the actual log output
>which would show what is going on.  I just tried a quick check to be sure
>it still works and all is well.  I don't remember exactly what caused this
>bug, but someone else found and reported it a good while back.  17.10 won't
>do this.


  Actually the version I have is 17.10, I downloaded it a little over a week
ago, from the link provided by timm manns web site. I will send you the log
file.



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.