Not logged inRybka Chess Community Forum
Forum RybkaChess Help Search Register Login
Up Topic Rybka Support & Discussion / Rybka Discussion / Rybka sees mate for black, but white wins next move!
- - By Fulcrum2000 (****) [nl] Date 2007-03-12 20:21 Edited 2007-03-12 20:24
Please have a look at this position:

r1br2k1/1pp2p2/6pp/P1P5/Q2nqB2/5NP1/P3pKBP/1R5R b - - 0 25


Rybka sees a mate for black and plays e1Q+, but white of course plays Rhxe1 (move 25/26) and the evaluation of rybka changes from mate for black to +5 for white... Strange indeed. I think this topic needs the 'bug' image so Vas can have a look at it.

Total game:
[Event "Rated game, 3m + 0s"]
[Site "Engine Room"]
[Date "2007.03.12"]
[Round "?"]
[White "OlegGR, HIARCS 11.1 MP UCI."]
[Black "Fulcrum2000, Rybka 2.3.1 mp"]
[Result "1-0"]
[ECO "A09"]
[WhiteElo "2761"]
[BlackElo "2787"]
[Annotator "0.21;0.23"]
[PlyCount "85"]
[EventDate "2007.03.12"]
[TimeControl "180"]

{HIARCS 11.1 MP UCI: 13.1 ply; 1.036kN/s Intel(R) Core(TM)2 CPU          6400
@ 2.13GHz 14MHz, (2 threads)} 1. Nf3 {0} d5 {B/0 0} 2. c4 {0} d4 {B/0 0} 3. g3
{0} g6 {B/0 0} 4. b4 {0} Bg7 {B/0 0} 5. d3 {0} e5 {B/0 0} 6. Bg2 {0} Nd7 {
0.23/13 6} 7. c5 {(Nbd2) 0.21/12 6} a5 {(Ne7) -0.04/14 5} 8. Qa4 {-0.01/12 8}
Ne7 {(Qe7) -0.15/14 0} 9. Na3 {(0-0) 0.04/13 19} O-O {-0.08/14 6} 10. Bg5 {
(0-0) 0.10/13 0} Qe8 {(c6) -0.29/13 7} 11. Nh4 {0.03/12 9} h6 {-0.33/17 11} 12.
Bd2 {0.03/14 0} e4 {(Nf6) -0.39/14 8} 13. dxe4 {0.36/12 9} d3 {-0.51/14 0} 14.
Rb1 {(Rc1) 0.21/12 7} dxe2 {-0.62/13 5} 15. f4 {(b5) 0.21/12 0} Nc6 {
(g5) -0.79/11 4} 16. Nc2 {(Nb5) 0.12/11 6} Qe7 {(Qe6) -0.96/11 6} 17. Ne3 {
0.41/11 6} Nd4 {-0.39/15 15} 18. Nd5 {0.10/13 9} Qe6 {(Qd8) -0.28/15 4} 19. e5
{(f5) 0.84/12 10} Nxe5 {(Bxe5) -0.63/11 17} 20. fxe5 {0.25/12 11} Bxe5 {
-0.54/15 1} 21. Nf4 {(bxa5) 0.00/12 9} Bxf4 {-0.59/10 4} 22. Bxf4 {0.69/11 2}
Qe7 {(Qc4) -0.75/12 0} 23. bxa5 {1.08/11 5} Rd8 {(Qxc5) -0.73/13 7} 24. Nf3 {
-0.56/11 14} Qe4 {(Nxf3+) -1.58/17 0} 25. Kf2 {-0.98/12 6} e1=Q+ {(Bh3) -#7/3 0
} 26. Rhxe1 {(Rbxe1) 5.38/11 2} Qc6 {4.40/13 2} 27. Qxc6 {5.22/12 0} Nxc6 {
4.55/12 0} 28. Bxc7 {5.28/11 2} Rd7 {4.65/14 0} 29. Bd6 {(Re8+) 5.41/11 1} Rd8
{4.69/13 1} 30. Ne5 {5.57/12 0} Nxe5 {4.81/14 5} 31. Rxe5 {5.83/13 0} Rxa5 {
4.96/13 1} 32. Bd5 {6.03/12 0} g5 {5.17/12 2} 33. Rb2 {6.07/11 0} Be6 {
(Kg7) 5.22/12 1} 34. Bxe6 {6.35/11 1} fxe6 {5.46/14 0} 35. Rxe6 {6.45/12 0} Rd7
{5.27/12 0} 36. Rxh6 {6.56/12 1} Kg7 {5.39/13 1} 37. Re6 {6.62/12 0} Kf7 {
5.38/12 1} 38. Re4 {7.07/11 0} Kf6 {(b6) 5.61/11 1} 39. Kf3 {(Kg2) 7.46/11 1}
Kg6 {(Kf5) 5.39/11 2} 40. Kg4 {7.87/11 1} Ra8 {6.69/13 1} 41. Re6+ {8.15/12 0}
Kf7 {7.18/14 0} 42. Kf5 {(Rh6) 8.26/12 0} g4 {(Re8) 8.13/15 9} 43. Rb4 {
(Rbe2) Fulcrum2000,Rybka 2.3.1 mp resigns 9.41/12 1} 1-0
Parent - - By Felix Kling (Gold) [de] Date 2007-03-12 20:33
Strange.

My Rybka 2.3.1 displays

Neue Partie, Blitz:3'
r1br2k1/1pp2p2/6pp/P1P5/Q2nqB2/5NP1/P4KBP/1R2q2R w - - 0 1


Analysis by Rybka 2.3.1 mp 32-bit :

2.Sxe1
  -+  (-62.89)   Tiefe: 1   00:00:00

(,  12.03.2007)


after e1Q and stops calculating after depth 1... I can't reproduce this with 2.3
Parent - - By Fulcrum2000 (****) [nl] Date 2007-03-12 20:46
I posted the wrong position first (after blacks move). Can you please try with the new diagram (before promotion)?.

Indeed 2.3 does not have this problem.
Parent - - By Felix Kling (Gold) [de] Date 2007-03-12 20:52 Edited 2007-03-12 20:57
I now created a position where Rybka gives a strange variation:

Neue Partie - Rybka 2.3.1 mp 32-bit
r1br2k1/1pp2p2/4q1pp/P1P5/Q2nqB2/5NP1/P3pKBP/1R5R b - - 0 1


Analysis by Rybka 2.3.1 mp 32-bit :

1...e1D+ 2.Tbxe1 Dxe1+ 3.Txe1 Dxe1+ 4.Kxe1 Sxf3+ 5.Lxf3
  +-  (#5)   Tiefe: 1   00:00:00
1...e1D+ 2.Tbxe1 Dxe1+ 3.Txe1 Dxe1+ 4.Kxe1 Sxf3+ 5.Ke2
  +-  (#5)   Tiefe: 1   00:00:00
1...e1D+ 2.Tbxe1 Dxe1+ 3.Txe1 Dxe1+ 4.Kxe1 Sxf3+ 5.Ke2 Sg1+ 6.Ke3
  +-  (#6)   Tiefe: 1   00:00:00
1...e1D+ 2.Tbxe1 Dxe1+ 3.Txe1 Dxe1+ 4.Sxe1 g5 5.a3
  +-  (#5)   Tiefe: 1   00:00:00
1...e1D+ 2.Tbxe1 Dxe1+ 3.Txe1 Dxe1+ 4.Sxe1 g5 5.Le5 Te8 6.Dxe8+ Kh7 7.Dxf7#
  +-  (#6)   Tiefe: 2   00:00:00
1...e1D+ 2.Tbxe1 Dxe1+ 3.Txe1 Dxe1+ 4.Sxe1 g5 5.Le5 Te8 6.Dxe8+ Kh7 7.Dxf7#
  +-  (#6)   Tiefe: 3   00:00:00
1...e1D+ 2.Tbxe1 Dxe1+ 3.Txe1 Dxe1+ 4.Sxe1 g5 5.Le5 Te8 6.Dxe8+ Kh7 7.Dxf7#
  +-  (#6)   Tiefe: 4   00:00:01
1...e1D+ 2.Tbxe1 De2+ 3.Txe2 Dxe2+ 4.Kg1 Sxf3+ 5.Lxf3 De1+ 6.Kg2 Td2+ 7.Lxd2 Dxd2+ 8.Kf1 Dc1+ 9.Ke2 Db2+ 10.Kd3 Lf5+ 11.Kc4 De2+ 12.Kd4 Td8+ 13.Dd7 Dd3+ 14.Ke5
  +-  (#14)   Tiefe: 5   00:00:02
1...e1D+ 2.Thxe1 Dxe1+ 3.Sxe1 De2+ 4.Kg1 Df2+ 5.Kxf2 Txa5 6.Sc2 Txa4 7.Le5 Sxc2 8.Td1 Txa2 9.Txd8+ Kh7 10.Th8#
  +-  (#9)   Tiefe: 6   00:00:03  297kN
1...e1D+ 2.Thxe1 Dxe1+ 3.Txe1 Dxe1+ 4.Sxe1 g5 5.Ld2 Se2 6.Lc3 Te8 7.Dxe8+ Kh7 8.Dxf7#
  +-  (#7)   Tiefe: 7   00:00:04  610kN
1...e1D+ 2.Thxe1 Dxe1+ 3.Txe1 Dxe1+ 4.Sxe1 g5 5.Le5 Sf3 6.Lc3 Sxe1 7.Kxe1 Te8+ 8.Dxe8+ Kh7 9.Dxf7#
  +-  (#8)   Tiefe: 8   00:00:06  884kN
1...e1D+ 2.Thxe1 Dxe1+ 3.Txe1 Dxe1+ 4.Sxe1 g5 5.Le5 Sf3 6.Lc3 Sxe1 7.a3 Sd3+ 8.Ke3 Te8+ 9.Dxe8+ Kh7 10.Dxf7#
  +-  (#9)   Tiefe: 9   00:00:11  1642kN
1...e1D+ 2.Thxe1 Dxe1+ 3.Txe1 Dxe1+ 4.Sxe1 g5 5.Le5 Sf3 6.Lc3 Sxe1 7.Ld5 Sd3+ 8.Kg1 Txd5 9.De8+ Kh7 10.Dxf7#
  +-  (#9)   Tiefe: 10   00:00:14  1960kN

(,  12.03.2007)
Parent - - By peter2804 (**) [gb] Date 2007-03-13 02:08
My 32 bit 2.3.1 never considers queening the pawn but goes for Bh3 and stays with it.
Parent - By M ANSARI (*****) [kw] Date 2007-03-13 06:15
Yes this is a good example of Rybka 2.3.1 promotion bug.  I think Vas is aware of it.  Apparentely what happens is that Rybka sees a mating variation after the promotion and simply stops searching all the possible moves of the opponent, thus it will not see moves that will not be mate.  I think that is an easy bug to take care off and should be fixed with 2.3.2.
Parent - By Fulcrum2000 (****) [nl] Date 2007-03-13 17:09
Some background information: I use Rybka 2.3.1 mp on an E6600. In multi-PV mode (2 or 3 variants) it considers playing the Queening option with a mate in x. Where x is not stable and changes when scrolling back and fort through the moves.
Parent - By grolich (***) Date 2007-03-13 14:19
partially Reproducible (in a variant of the problem):
same position, but different : Rybka on my computer does suggest Bh3. But if queening is PLAYED, it thinks that white is LOSING after its ONLY suggested move (with no further line): NxQ?? (value: -293.52, strange value for Rybka to show when it does not show mate in ... I got used to those numbers from crafty in the past... Never from Rybka).
Tried removing multi PV, change to outlook to ultra optimistic and back to normal. Same behavior.

Only after RxQ is played does Rybka get more "sensible" in its outlook, and understands that white is winning.

Actually even after the horrible NxQ it sees that it is +3.9 better... not the strange (enormous and variationless) value that appeared before...

The only difference is that sometimes after queening it thinks of RxQ but still gives the same strange value
(tends to happen in multi PV mode), when it does, it actually gives a (very) short line: (On my system, it shows Rbe1 as first...

FEN:
r1br2k1/1pp2p2/6pp/P1P5/Q2nqB2/5NP1/P4KBP/1R2q2R w - - 0 26


Rybka 2.3.1 32-bit:
---------------------------------------------------------------------------
   1  00:00            10  10.240  -293.52  Rh1xe1
   1  00:00             9  9.216  -293.52  Nf3xe1
   1  00:00             9  9.216  -293.52  Rb1xe1 Qe4d3 Nf3xd4 Rd8xd4 Qa4e8+

As you can see, it does not think about looking too deeply on these lines... At least it gives one here, even if a strange one.
This output is with Multi PV = 4 mode.

I'm used Arena for this test. The tests were made in ALL of the outlook settings (pessimistic to optimistic), with multi PV on AND with multi PV off. Same results.

Rybka 2.3.1 32bit sp.

Athlon 3400+
Parent - By Vasik Rajlich (Silver) [hu] Date 2007-03-14 12:56
Hi,

this is a known bug in 2.3.1, will be fixed in 2.3.2.

Vas
Up Topic Rybka Support & Discussion / Rybka Discussion / Rybka sees mate for black, but white wins next move!

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill