Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Rybka Discussion / strange rybka behaviour in pawnendgame
- - By rivaldo (***) Date 2007-12-14 02:50
I recently played following game as black. my opponent was much weaker, but I lost the interesting game as you can see below.
the reason I am psoting this is the possible pawnendgame with 4 pawns each, which I calculated pretty accurately in timetrouble, but wasn't sure it is won and alas rejected it.
when analysing the endgame with rybka 2.32a I already guessed, that this tactical endgame would be a problem for computers. rybka "understands" the ideas rather late but more important her behaviour is a bit strange. there is a huge difference whether you analyse it single or multi pv mode and at a certain point it won't reach the next depth. if you give the right moves until she sees the point, she will remember it more or less, if you go back. but there are positions, where she doesn't see the black win in many minutes and one move later she sees it in seconds. is this normal? if it's normal she shouldn't cut off these lines in such a tactical endgame.
anyway this endgame is fun and maybe future versions are able to see deeper or understand it better.

[Site "?"]
[Date "2007.12.09"]
[Round "3"]
[White "opponent"]
[Black "rivaldo"]
[Result "1-0"]
[ECO "D44"]
[PlyCount "84"]

1. d4 d5 2. c4 c6 3. Nc3 Nf6 4. Nf3 e6 5. Bg5 dxc4 6. e4 b5 7. e5 h6 8. Bh4 g5
9. Nxg5 hxg5 10. Bxg5 Nbd7 11. exf6 Qb6 12. g3 Bb7 13. Bg2 c5 14. d5 O-O-O 15.
O-O b4 16. Ne2 $146 Ne5 (16... exd5) (16... Bh6) (16... Bxd5 17. Bxd5 Ne5 18.
Nf4 exd5) 17. Nf4 Bh6 18. Bxh6 (18. Qh5) 18... Rxh6 19. Qe2 Nd3 20. Nxd3 cxd3
21. Qxd3 (21. Qe3) 21... Bxd5 22. Qc2 Bxg2 23. Kxg2 Qc6+ 24. f3 Rxf6 25. a3 a5
26. Qe2 Rf5 27. axb4 axb4 28. Ra6 Qb7 29. Rfa1 Rfd5 30. Ra8+ Kc7 31. R8a7 Rd2
32. Rxb7+ Kxb7 33. Kf1 Rxe2 34. Kxe2 Kc6 35. b3 $6 f5 36. Rd1 $2 (36. Ra6+ Kd5
37. Ra2) 36... Rxd1 $3 (36... Rh8 {game continuation} 37. h4 e5 38. Rh1 c4 39.
bxc4 Kc5 40. Kd3 Rd8+ 41. Kc2 Kxc4 42. h5 b3+ 43. Kb1 Kc3 44. Rc1+ Kd2 45. h6
Rh8 46. Rh1 Ke3 $2 (46... e4 47. fxe4 fxe4 48. g4 e3 49. g5 e2 50. g6 Rxh6 $19
{I missed Rxh6 and threw away the win}) 47. g4 Kf4 48. gxf5 {with draw offer}
Kxf3 $4 {still playing for the win without thinking} 49. f6 e4 50. f7 e3 51.
Rf1+ (51. f8=Q+ Rxf8 52. h7 e2 53. h8=Q Rxh8 $19 {was the thing I had in mind})
) 37. Kxd1 e5 38. h4 (38. Kd2 Kd5 39. Kd3 (39. Ke3 c4 $19) (39. Kc2 e4 $19) (
39. h4 Ke6 40. Kd3 Kf6 41. Kd2 c4 42. bxc4 e4 43. fxe4 fxe4 44. c5 b3 45. c6
Ke6 $19) 39... e4+ 40. fxe4+ fxe4+ 41. Kd2 c4 $19) 38... Kd5 39. h5 (39. Kd2
Ke6 40. g4 (40. Kd3 Kf6 41. Kc4 (41. Kd2 Kf7 42. Kd3 Kg6 43. Kd2 Kh5 44. Kd3 f4
45. gxf4 exf4 46. Ke4 Kxh4 47. Kxf4 c4 48. Ke3 c3 $19) 41... e4 42. fxe4 fxe4
$19 {this is the basic idea behind this endgame. since I saw it, I knew I
couldn't lose it, but I rejected Rxd1, because I wasn't 100% sure, if it's won.
}) 40... fxg4 41. fxg4 c4 42. bxc4 e4 43. h5 b3 44. h6 Kf7 $19) 39... Ke6 40.
Kd2 Kf6 41. g4 (41. Ke2 c4) (41. Kd3 Kg5 42. g4 e4+ $19) 41... e4 42. fxe4 fxg4
$19 1-0
Parent - By Vasik Rajlich (Silver) Date 2007-12-16 16:15
Yes, what you describe is certainly normal. It's really the main reason why human + computer are stronger than computer alone, the human can guide the search in critical moments like these.

Vas
Up Topic Rybka Support & Discussion / Rybka Discussion / strange rybka behaviour in pawnendgame

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill