Author: Brian Richardson
Date: 07:03:00 10/29/01
Go up one level in this thread
On October 29, 2001 at 05:37:15, Ed Schröder wrote:
>On October 29, 2001 at 01:30:12, Dieter Buerssner wrote:
>
>>Tinker caught Yace in a nice stalemate trap.
>>
>>[D] 7k/6R1/6R1/1r5P/6K1/8/8/8 b - - 0 55
>>
>>Yace just captured on g7. It is easy to see for a human, but engines seem to
>>have problems with this. Yace got a nice fail low from +7.4 to draw score in the
>>next move. Do other engines see this immediately? With and without TBs (KRRKR
>>can help here). I have some stalemate detection code in quiescence search, which
>>is normally disabled. I enabled this again, in the hope, that this will make it
>>much faster. But with no success.
>
>Nice position... I tried to find the longest (is it?) defence:
>
>1..Rg5 2.Kf4 Rf5 3.Ke4 Re5 4.Kd4 Rd5 5.Kc4 Rc5 6.Kb4 Rb5 7.Kc3 Rc5
>8.Kd3 Rd5 9.Ke3 Re5 10.Kf3 Rf5 11.Kg3 Rg5 12.Kf2 Rf5 13.Ke2 Re5 14.Kd2
>Rd5 15.Kc2 Rc5 16.Kb2 Rb5 17.Kc1 Rc5 18.Kd1 Rd5 19.Ke1 Re5 20.Kf1 Rf5
>21.Kg2 Rg5 22.Kh3 Rg3 23.Kh2 Rg2 24.Kh1 Rg1 25.Kh2 Rg2 -> repetition
>
>Typical position for the King I would say.
>
>But how can you evaluate (detect) stalemate in QS?
>
>Looks pretty dangerous to me because of its selective nature and
>you can't just say score=0.00 in case no more captures or checks
>are generated. So what is the big secret?
>
>Ed
>
Tinker has a draw heuristic it uses as part of the repetition test in both full
width and q search. It finds this draw with or without EGTBs in about 10 sec at
ply 14 (with EGTB) or 12 (without), plus extensions. It does not actually look
for stalemate, but "perpetual" check instead. Not exactly sure how it worked in
this case. I recently modified the code to require more checks to trigger the
draw condition.
Brian
>
>
>
>>Here is the whole game:
>>
>>[Event "Computer chess game 0 300 +3 0 0"]
>>[Site "ICC"]
>>[Date "2001.10.28"]
>>[Round "5"]
>>[White "Yace 0.99.57"]
>>[Black "Tinker"]
>>[Result "1/2-1/2"]
>>[WhiteElo "2972"]
>>[BlackElo "2519"]
>>[ECO "B78"]
>>[Opening "Sicilian"]
>>[Variation "Dragon, Yugoslav attack, 10.O-O-O"]
>>[Timecontrol "0 300 +3 0 0"]
>>[Time "Sun Oct 28 01:07:14 2001"]
>>[last_book_move "17. f5"]
>>
>>1. e4 {0s} c5 {1s} 2. Nf3 {0s} d6 {2s} 3. d4 {0s} cxd4 {1s} 4. Nxd4 {0s} Nf6
>>{2s} 5. Nc3 {0s} g6 {2s} 6. Be3 {0s} Bg7 {2s} 7. f3 {0s} O-O {2s} 8. Qd2 {0s}
>>Nc6 {2s} 9. Bc4 {0s} Bd7 {2s} 10. O-O-O {0s} Rc8 {2s} 11. Bb3 {0s} Ne5 {2s}
>>12. h4 {0s} h5 {3s} 13. Bg5 {0s} Rc5 {2s} 14. g4 {0s} hxg4 {2s} 15. f4 {0s}
>>Nc4 {2s} 16. Qe2 {0s} b5 {2s} 17. f5 {0s} Qc8 {2s} 18. fxg6 {0.48/8 33s} a6
>>{2s} 19. gxf7+ {0.53/7 25s} Kxf7 {2s} 20. Bxc4+ {0.54/8 20s} Rxc4 {0s} 21.
>>Rhf1 {0.58/7 18s} Rxc3 {0s} 22. bxc3 {0.58/8 10s} Qxc3 {29s} 23. Qd3
>>{0.62/9 0s} Qxd3 {0s} 24. cxd3 {0.55/10 14s} Kg6 {0s} 25. Ne2 {0.77/9 14s} e5
>>{0s} 26. Kd2 {0.74/9 10s} d5 {9s} 27. Bxf6 {0.45/9 3s} Rxf6 {1s} 28. exd5
>>{0.50/10 7s} Rd6 {1s} 29. Nc3 {0.51/10 9s} b4 {16s} 30. Ne4 {0.44/11 0s} Rxd5
>>{9s} 31. h5+ {0.45/10 4s} Kh6 {0s} 32. Rf7 {0.49/11 11s} Be6 {0s} 33. Rc7
>>{1.13/10 13s} Bf5 {0s} 34. Rc6+ {1.15/10 7s} Kh7 {15s} 35. Ng5+ {1.17/10 0s}
>>Kg8 {5s} 36. Ke3 {1.17/10 3s} g3 {7s} 37. Rxa6 {1.67/9 1s} g2 {5s} 38. Ne4
>>{2.17/10 7s} b3 {0s} 39. axb3 {2.99/9 7s} Bxe4 {8s} 40. Kxe4 {3.67/11 4s} Rd4+
>>{5s} 41. Ke3 {3.73/12 7s} Rg4 {0s} 42. Rg1 {3.87/11 12s} Rg3+ {0s} 43. Ke4
>>{4.75/11 11s} Kh7 {15s} 44. Rg6 {4.75/12 0s} Rh3 {8s} 45. R1xg2 {4.77/11 3s}
>>Bh6 {8s} 46. Kxe5 {4.82/10 3s} Rxd3 {0s} 47. Rc2 {4.94/9 11s} Bg7+ {6s} 48.
>>Ke4 {4.97/10 5s} Rd4+ {5s} 49. Ke3 {5.12/11 1s} Rd7 {8s} 50. b4 {5.29/10 3s}
>>Re7+ {7s} 51. Kf3 {5.44/12 3s} Rf7+ {7s} 52. Kg4 {5.43/12 2s} Rb7 {0s} 53. b5
>>{5.76/11 11s} Rxb5 {0s} 54. Rc7 {7.45/12 9s} Kh8 {1s} 55. Rcxg7 {7.43/11 4s}
>>Rg5+ {7s} 56. Kf3 {0.00/18 1s} Rf5+ {0s} 57. Ke2 {0.00/21 6s} Re5+ {0s} 58.
>>Kd1 {0.00/23 8s} Rd5+ {6s} 59. Ke1 {0.00/24 4s} Re5+ {0s} 60. Kd1 {0.00/25 6s}
>>Rd5+ {7s} 61. Kc2 {0.00/25 3s} Rc5+ {0s} 62. Kd2 {0.00/26 10s} Rd5+ {0s} 63.
>>Ke3 {0.00/26 10s} Re5+ {0s} 64. Kd4 {0.00/26 9s} Rd5+ {5s} 65. Kxd5
>>{0.00/27 2s} 1/2-1/2 {Tinker stalemated}
>> <snip>
>>
>>Regards,
>>Dieter
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.