Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Rybka Discussion / Bug in Rybka 2.3.2a ?
- - By Vinvin (***) Date 2008-01-02 02:20
It's seems to be a problem related to move history :
Load the following game and analyse the 44th white move with 3-best moves , you'll get this :

Analysis by Rybka 2.3.2a mp 32-bit :

1. +-  (1.79): 44.Ta8 Te7 45.Td8 h6 46.Td6 Te6 47.Td7 Fb8 48.Txd4 Tc6 49.Td7 Fd6
2. +-  (1.75): 44.Tc8 Rf6 45.Cf4 Td6 46.Re2 Re7 47.Rd3 Td7 48.Th8 Tb7 49.Txh7 Fb8 50.Cd5+ Re6
3. =  (0.00): 44.Td8 Rf6

Why 44.Rd8 get this 0.00 ???
When you play 44.Rd8 on the board , Rybka says 44...Kf6 0.00 but when you play Kf6 on the board there's no more 0.00 ! very Strange behaviour !

Best whishes, Vincent.

Here's the game :

[Event "16m + 0s, rated"]
[Site "?"]
[Date "2008.01.02"]
[Round "?"]
[White "Milde, Rybka 2.3.2a mp"]
[Black "Nemrod, Rybka 2.3.2a mp"]
[Result "*"]
[WhiteElo "2644"]
[BlackElo "2576"]
[PlyCount "111"]
[EventDate "2008.01.02"]
[EventType "rapid"]

1. d4 {B/0 0} d5 {B/0 0} 2. c4 {B/0 0} c6 {B/0 0} 3. Nc3 {B/0 0} e6 {B/0 0} 4.
e3 {B/0 0} Nf6 {B/0 0} 5. Nf3 {B/0 0} Nbd7 {B/0 0} 6. Qc2 {B/0 0} Bd6 {B/0 0}
7. Bd3 {B/0 0} O-O {B/0 0} 8. O-O {B/0 0} dxc4 {B/0 0} 9. Bxc4 {B/0 0} b5 {
B/0 0} 10. Be2 {B/0 0} Bb7 {B/0 0} 11. Rd1 {B/0 0} Qc7 {B/0 0} 12. e4 {B/0 0}
e5 {B/0 0} 13. Bg5 {B/0 0} exd4 {0.04/14 20} 14. Rxd4 {B/0 0} Ne5 {0.03/14 9}
15. Bxf6 {B/0 0} gxf6 {0.04/16 16} 16. h3 {(g3) B/0 0} Rad8 {-0.10/15 32} 17.
Nh4 {(Tdd1) 0.02/17 11} b4 {0.01/16 27} 18. Na4 {0.08/20 8} c5 {0.00/16 1} 19.
Rdd1 {0.06/19 11} Nc6 {0.08/16 2} 20. Nf5 {0.05/19 16} Nd4 {0.10/17 3} 21. Nxd4
{0.10/19 6} cxd4 {0.11/17 6} 22. Qxc7 {0.18/20 7} Bxc7 {0.10/18 5} 23. Nc5 {
0.17/20 2} Bc6 {0.11/17 10} 24. Rac1 {0.21/20 7} Bb6 {0.28/17 27} 25. Bd3 {
(Ce6) 0.29/21 7} Rfe8 {0.18/18 22} 26. Na6 {0.29/21 22} Bxe4 {0.26/18 12} 27.
Bxe4 {0.33/21 40} Rxe4 {0.28/20 21} 28. Nxb4 {0.38/21 8} Re2 {0.29/18 67} 29.
Nd3 {0.42/20 6} f5 {0.38/18 31} 30. Kf1 {0.54/21 21} Rde8 {0.47/17 5} 31. Re1 {
(a4) 0.59/20 17} R2e4 {0.36/16 8} 32. b4 {0.65/20 9} Kg7 {0.46/15 10} 33. a4 {
(Ted1) 0.74/20 29} R8e7 {0.62/15 21} 34. Red1 {0.93/19 30} a6 {0.84/18 7} 35.
a5 {(Tc6) 1.40/23 24} Ba7 {1.19/21 24} 36. Rc6 {1.56/24 12} R4e6 {1.42/21 40}
37. Rdc1 {(Txe6) 1.58/25 13} Kf6 {1.60/20 33} 38. Rxe6+ {1.70/24 27} Rxe6 {
1.59/23 35} 39. Rc8 {1.67/26 78} Re4 {1.69/24 20} 40. Rh8 {(f4) 1.87/22 13} Kg7
{1.81/22 55} 41. Rd8 {1.99/26 36} Kf6 {1.81/23 46} 42. g3 {2.26/26 33} Re6 {
1.73/21 25} 43. Rh8 {2.37/25 8} Kg7 {1.80/21 5} 44. Ra8 {2.53/24 20} Re7 {
2.22/21 21} 45. Rd8 {2.56/24 0} Kf6 {2.57/21 60} 46. Nf4 {2.67/25 21} Ke5 {
2.67/21 11} 47. Ke2 {2.77/24 46} d3+ {2.58/22 0} 48. Nxd3+ {2.77/24 68} Kf6+ {
2.70/25 0} 49. Kd2 {(Rf3) 2.77/24 54} Re6 {2.67/18 12} 50. Rd7 {3.55/24 28} Bb8
{3.40/23 0} 51. Nc5 {3.55/22 1} Rc6 {3.41/21 5} 52. Kd3 {(Tb7) 3.55/21 4} Kg7 {
3.53/15 6} 53. Rb7 {4.39/20 11} Bd6 {3.87/17 0} 54. Kd4 {4.67/19 8} Bxc5+ {
4.05/16 1} 55. bxc5 {4.93/18 7} Rc8 {4.05/17 2} 56. Rb6 {4.93/17 6} *
Parent - By FirebrandX (**) Date 2008-01-02 02:50
I noticed similar issues with Rybka on what I call "3-fold prospects", where it will eval a move that seemingly invites 3-fold with a 0.00, when actually one side or the other is losing. I've found I have to copy the position and re-paste it so Rybka doesn't know the move history and thus will evaulate the position fairly without being clouded by thoughts of a 3-fold.
Parent - - By Vempele (Silver) Date 2008-01-02 10:00
This is well-known. Rybka, like most engines, considers 2-fold repetitions drawn.
Parent - - By BB (****) Date 2008-01-02 10:13
like most engines, considers 2-fold repetitions drawn

I would be interested in a list of which engines do/don't. Certainly it is much easier simply to implement 2-fold and forget it, but are (say) the majority of commercial engines this indifferent about repetition?
Parent - By Permanent Brain (*****) Date 2008-01-02 10:55
Shredder 9 SE, after depth 18:

1. +-  (2.63): 44.Ra8 Re7 45.Rd8 h6 46.Rd6 Re6 47.Rxe6 fxe6 48.Nc5 Bxc5 49.bxc5 Kf7 50.f4 Ke7 51.Ke2 Kd7 52.Kd3 Kc6 53.Kxd4 Kb5 54.Ke5 Kxc5 55.Kxe6
2. +-  (2.47): 44.Rd8 Rc6 45.Rd7 Bb8 46.Rxd4 Kf6 47.Rd5 Ba7 48.b5 axb5 49.Rxb5 Rc3 50.Ke2 Rc2+ 51.Kf3 Rd2 52.Rb3 Ra2 53.Rb5
3. +-  (2.13): 44.Rc8 Kf6 45.Rd8 Rc6 46.Rd7 Bb8 47.Rxd4 Bd6 48.Ke2 h6 49.Ke3 Ke6

I think Shredder checks if the repetition is forced/forceable.

But Fritz (at least up to version 8.1.3.0 = Fritz for Fun 4), after depth 15:

1. +-  (2.09): 44.Ra8 Re7 45.Rd8 h6 46.Rd6 Re6 47.Rd7 Bb8 48.Nc5 Rc6 49.Rxd4 Be5 50.Rd3 Rg6 51.Rd8
2. +-  (1.90): 44.Rc8 Kf6 45.Nf4 Rd6 46.Ke2 d3+ 47.Nxd3 Bd4 48.Nc5 Ke5 49.Rc7 Kf6 50.Nd7+ Ke7 51.Nb8+ Kf6 52.Nc6
3. =  (0.00): 44.Rd8 Kf6
Parent - By Vempele (Silver) Date 2008-01-02 11:06

> Certainly it is much easier simply to implement 2-fold and forget it, but are (say) the majority of commercial engines this indifferent about repetition?


I've heard Shredder has 3-fold. Fruit 2.1 doesn't, Toga has code for it but it's commented out. Zappa 1.1 even ignored a 3-fold repetition when I tested it. Junior uses at least one trick to forward prune some moves that would lead to repetition, but I don't know if it uses 2-fold or 3-fold.
Parent - By Vinvin (***) Date 2008-01-02 19:22
It's hard to find where engines "see" a 2- fold repetition.
Up Topic Rybka Support & Discussion / Rybka Discussion / Bug in Rybka 2.3.2a ?

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill