Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Rybka Discussion / Error reading tablebases in 2.3.x
- - By Graham Burgess Date 2007-07-26 11:25
I noticed something odd with Rybka's reading of the tablebases in the following R+P vs R+P position:

8/6kp/1K6/1PR5/8/8/8/5r2 w - - 0 55


It's the position after Black's 54th move in the game Macieja-Gershon, Paget Parish 2001 (full PGN given at the end of this post).

In its (sorry, oops, I mean her) analysis window, Rybka 2.3.2a 32-bit gives only 55 Kb7 as winning, and all other moves get the odd evaluation 0.05. This includes several other winning moves, and the two obviously losing rook blunders. Once one of the other winning moves (such as 55 Kc7) is played, it is correctly assessed (all Black's replies given as losing). It is a similar story in 2.3.2, except that two moves are given as winning. The last version where the tablebase appears to be read correctly in this position is 2.2n2.

I am using the single-processor version of Rybka as a UCI engine via ChessBase 9. The version of Rybka 2.3.2a that I am using is dated 26-June-2007.

[Event "Elbow Beach Club GM-A"]
[Site "Paget Parish"]
[Date "2001.01.30"]
[Round "10"]
[White "Macieja, Bartlomiej"]
[Black "Gershon, Alik"]
[Result "1-0"]
[ECO "B01"]
[WhiteElo "2578"]
[BlackElo "2540"]
[PlyCount "120"]
[EventDate "2001.01.20"]
[EventType "tourn"]
[EventRounds "10"]
[EventCountry "BER"]
[EventCategory "13"]
[Source "ChessBase"]
[SourceDate "2001.03.20"]

1. e4 d5 2. exd5 Nf6 3. Nf3 Bg4 4. Bb5+ Nbd7 5. h3 Bh5 6. Nc3 a6 7. Be2 Nb6 8.
d4 Nfxd5 9. Nxd5 Qxd5 10. O-O e6 11. b3 Be7 12. c4 Qd8 13. Be3 O-O 14. Ne5 Bxe2
15. Qxe2 Bf6 16. Rad1 Qe7 17. Qf3 Rab8 18. Nd3 Bg5 19. Bxg5 Qxg5 20. Nc5 c6 21.
Rd3 Rfd8 22. Rfd1 Nd7 23. Ne4 Qe7 24. R1d2 Nf6 25. Nxf6+ Qxf6 26. Qxf6 gxf6 27.
d5 cxd5 28. cxd5 e5 29. d6 Rd7 30. Rd5 Rc8 31. f4 Rc6 32. Kf2 Kg7 33. Kf3 Rb6
34. fxe5 fxe5 35. Ke4 f6 36. Kf5 Kf7 37. Rc2 Rbxd6 38. Rxd6 Rxd6 39. Rc7+ Kg8
40. Rxb7 Rd2 41. Kxf6 Rf2+ 42. Kxe5 Re2+ 43. Kd4 Rxa2 44. Ra7 Rb2 45. Kc4 Rc2+
46. Kb4 Rxg2 47. Rxa6 Rg3 48. h4 Rh3 49. Rc6 Rxh4+ 50. Kc5 Rh5+ 51. Kb6 Rh1 52.
b4 Kg7 53. b5 Rb1 54. Rc5 Rf1 55. Kc7 Rf7+ 56. Kb8 Rf8+ 57. Ka7 Rf7+ 58. Ka6
Kg6 59. b6 h5 60. b7 Rf6+ 1-0
Parent - By sylvain.renard.27 (**) Date 2007-07-26 14:02
Hello,
I have the same (wrong) results with another GUI (Schredder Classic 2.1).
The ToDo list of Vas is growing :)
  Best regards,
     Sylvain
Parent - By Caissafan (***) Date 2007-07-27 09:19
Hi,

I got this under CA:

[White "Rybka 2.3.2a mp"]
[Black ""]
[Remark "0:00.00-0:00.00"]
[SetUp "1"]
[FEN "8/6kp/1K6/1PR5/8/8/8/5r2 w - - 0 2"]

2.Kc6 Rf6+ 3.Kb7 h5 4.b6 Rf1 5.Ka6 Ra1+ 6.Ra5 Rc1 7.b7 Rc6+ 8.Kb5
Rc2 {(0:02.20) [+4.56] d=18} *

[White "Rybka 2.3.2a mp"]
[Black ""]
[Remark "0:00.00-0:00.00"]
[SetUp "1"]
[FEN "8/6kp/1K6/1PR5/8/8/8/5r2 w - - 0 2"]

2.Kc7 Rf7+ 3.Kc6 Rf6+ 4.Kb7 h5 5.b6 Rf1 6.Ka6 Ra1+ 7.Ra5 Rc1
8.b7 Rc6+ {(0:02.20) [+4.56] d=18} *
Parent - By Vasik Rajlich (Silver) Date 2007-07-28 08:24
Thanks.

It seems that this behavior is unique to 6 man TBs. If you remove the h7 pawn, everything is fine.

Anyway, it's something to fix for Rybka 3.

Vas
Up Topic Rybka Support & Discussion / Rybka Discussion / Error reading tablebases in 2.3.x

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill