Author: martin fierz
Date: 04:06:50 02/12/04
Go up one level in this thread
On February 12, 2004 at 06:03:37, Uri Blass wrote:
>On February 12, 2004 at 04:48:26, martin fierz wrote:
>
>>On February 11, 2004 at 20:40:17, scott farrell wrote:
>>
>>>this is from a game on ICC crafty vs chompsterx
>>>
>>>I think Robert is using his cct6 quad opteron, and chompster wins due to a
>>>blunder by crafty.
>>>
>>>in this position crafty plays rxh3, will a huge scores favouring black (crafty
>>>v19.3 analysis), and when white plays Rh2 it then changes the score to like +4
>>>for white, a swing of about a queen.
>>>
>>>It took crafty 2:40 (2 mins 40 secs) and 17 ply to see the fail low (on a dual
>>>PIII 1.26 v19.3).
>>>
>>>[d] 8/7p/8/4pp1k/8/r6P/P5R1/6K1 b - - 0 41
>>>
>>>
>>>[Event "ICC 5 3"]
>>>[Site "Internet Chess Club"]
>>>[Date "2004.02.11"]
>>>[Round "-"]
>>>[White "ChompsterX"]
>>>[Black "crafty"]
>>>[Result "1-0"]
>>>[ICCResult "Black resigns"]
>>>[WhiteElo "2572"]
>>>[BlackElo "2709"]
>>>[Opening "Bishop's opening: Berlin defense"]
>>>[ECO "C24"]
>>>[NIC "IG.04"]
>>>[Time "19:55:16"]
>>>[TimeControl "300+3"]
>>>
>>>1. e4 e5 2. Bc4 Nf6 3. d3 Nc6 4. Nc3 Bb4 5. Bd2 O-O 6. Nd5 Bxd2+ 7. Qxd2 Nd4
>>>8. c3 b5 9. cxd4 bxc4 10. dxe5 Nxd5 11. exd5 cxd3 12. Qxd3 d6 13. Nf3 Rb8
>>>14. O-O Rxb2 15. Qd4 Rb5 16. Qd3 Rb6 17. Rfe1 dxe5 18. Nxe5 Bb7 19. Rad1 Rd6
>>>20. Qa3 Bxd5 21. Qxa7 Bxg2 22. Rxd6 cxd6 23. Kxg2 Qg5+ 24. Kh1 dxe5 25. Qc5
>>>f6 26. Qd5+ Kh8 27. Rg1 Qf4 28. Qb7 g6 29. f3 Rd8 30. Qe4 Qxe4 31. fxe4 Rd3
>>>32. Re1 Rf3 33. Kg2 Ra3 34. Re2 Kg7 35. h3 Kh6 36. Kh2 Kg5 37. Rg2+ Kh4 38.
>>>Rg4+ Kh5 39. Rg2 f5 40. exf5 gxf5 41. Kg1 Rxh3 42. Rh2 Rxh2 43. Kxh2 e4 44.
>>>a4 Kg4 45. a5 e3 46. Kg2 f4 47. a6 f3+ 48. Kf1 Kg3 49. a7 e2+ 50. Ke1 h5 51.
>>>a8=Q h4 52. Qb8+ Kg4 53. Qc8+ Kf4 54. Qh3 Ke4 55. Qxh4+ Ke3 56. Qe7+ Kd3 57.
>>>Kf2 e1=Q+ 58. Qxe1 Kd4 59. Kxf3 Kc4 60. Qa5 Kb3 61. Kf4 Kb2 62. Qb4+ Ka1 63.
>>>Ke4 Ka2 64. Kf4 {Black resigns} 1-0
>>
>>for all those who immediately see that white wins after ...Rxh3?? Rh2 with their
>>passed pawn race code, check this minor modification (i gave black one tempo
>>more)
>>
>>[D]8/7p/8/5p1k/4p3/r6P/P5R1/6K1 b - -
>>
>>this time it's black who wins after ...Rxh3 Rh2. you might want to check that
>>position to see whether your pawn race code is correct in this case too.
>>
>>for me, this is another of these extremely sharp positions, where a single tempo
>>decides about win or loss - i suspect that many evals will be confused by this,
>>as it is basically only decidable by search. but perhaps you guys can prove me
>>wrong?
>>
>>cheers
>> martin
>
>The main point is that programs should not evaluate unclear position as winning.
well you're right about that of course, and it's one of the reasons i posted
this modified position. i would say that those who had no trouble finding that
....Rxh3 is a big mistake in the original position are probably lucky with their
pawn race eval, and might just make an equally large mistake in the second
position by still evaluating it as good for white.
of course all of this is *not* about what move you make after thinking for 5
seconds, but just about the static eval of the position after ...Rxh3 Rh2 Rxh2
Kxh2. that is extremly hard to evaluate, and i'd be surprised if anyone can do
this properly without a search.
in my engine, i also decide to basically return a "don't know" score if both
sides have dangerous passers. however, i'm not 100% sure whether you're not
better off trying to identify who wins, and hope that you get it right more
often than not - instead of not evaluating. i mean, it's clear to everybody that
this position is either a white win or a black win, so returning a close-to-zero
score is also rather far from the truth in both cases...
cheers
martin
>
>Black has a big advantage without capturing the pawn so there is no reason to
>capture h3 and get unclear pawn endgame.
>
>evaluating Rxh3 Rh2 as a win for white without search is a msitake but
>evaluating it as a win for black is also a mistake that cause Crafty to miss a
>simple win by not capturing the pawn.
>
>Uri
>Uri
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.