Author: Vincent Diepeveen
Date: 12:22:08 09/17/00
Go up one level in this thread
On September 16, 2000 at 14:27:33, Robert Hyatt wrote:
>On September 16, 2000 at 10:15:33, Vincent Diepeveen wrote:
>
>>Hello,
>>
>>During the WMCC after round #1 i was very sick DIEP allowed shredder
>>to exchange to a rook ending which appeared to be an easy win for
>>shredder, instead to an opposite bishops rook endgame.
>>
>>So i directly putted the penalty for opposite bishops higher!
>>
>>Now i analyzed just for fun the position where deep blue played 37.Be4?
>>against Deep Blue.
>>
>>
>>[Event "ICC u 120 0 05/04/97"]
>>[Site "Internet Chess Club"]
>>[Date "1997.05.04"]
>>[Round "-"]
>>[White "Deep-Blue"]
>>[Black "Kasparov"]
>>[Result "1-0"]
>>[WhiteElo "0"]
>>[BlackElo "2795"]
>>[GameID "862766392"]
>>[ECO "C93"]
>>[NIC "RL.22"]
>>[LongECO "Ruy Lopez: closed, Smyslov defense"]
>>
>> 1. e4 e5 2. Nf3 Nc6 3. Bb5 a6 4. Ba4 Nf6 5. O-O Be7
>> 6. Re1 b5 7. Bb3 d6 8. c3 O-O 9. h3 h6 10. d4 Re8
>>11. Nbd2 Bf8 12. Nf1 Bd7 13. Ng3 Na5 14. Bc2 c5 15. b3 Nc6
>>16. d5 Ne7 17. Be3 Ng6 18. Qd2 Nh7 19. a4 Nh4 20. Nxh4 Qxh4
>>21. Qe2 Qd8 22. b4 Qc7 23. Rec1 c4 24. Ra3 Rec8 25. Rca1 Qd8
>>26. f4 Nf6 27. fxe5 dxe5 28. Qf1 Ne8 29. Qf2 Nd6 30. Bb6 Qe8
>>31. R3a2 Be7 32. Bc5 Bf8 33. Nf5 Bxf5 34. exf5 f6 35. Bxd6 Bxd6
>>36. axb5 axb5 37. Be4 Rxa2 38. Qxa2 Qd7 39. Qa7 Rc7 40. Qb6 Rb7
>>41. Ra8+ Kf7 42. Qa6 Qc7 43. Qc6 Qb6+ 44. Kf1 Rb8 45. Ra6
>>{Black resigns} 1-0
>>
>>Now after move 37.Qb6 white is won. However to my big surprise DIEP's
>>score drops and drops when searching here. It drops to about 0.10 !!!!!!
>>
>>Really incredible. the reason appears to be very simple. After
>>37.Qb6 Rxa2 28.Rxa2,Ra8 29.Ra5,Rxa5 30.bxa5,Qb8 then black forces to
>>an opposite bishop ending! 31.Qxb8,Bxb8
>>
>>However, the bishop ending is lost for black. And exactly that's what DIEP
>>is misevaluating now. Only after i start making moves and move with the king
>>from g1 to g4 then it starts getting fail lows. Obviously white has
>>2 passed pawns which keep both the black king AND bishop busy, where blacks
>>2 potential passed pawns cannot form a real threat.
>>
>>32. Kf2,.. 33.Kf3 etcetera.
>>
>>However one can do all kind of weird moves with black before the
>>evaluator will see the problem!
>>
>>So where it's obvious that white has a real cool try to win this endgame,
>>and where black must do real effort to prevent losing it, there the long
>>time till white can force into a position which is evident for
>>the programs to win, it's logical that if you use real high scores for things
>>like opposite bishops, that Be4 gets played. the fact that Deep Blue gets
>>this fail high so low so soon (though score is much higher as the 0.10
>>from diep at 21 ply for Qb6) this explains easily that deep blue is based
>>upon real high scores.
>>
>>All the e4 lines are completely nonsense. No chance of getting those lines.
>>It's the opposite bishop score that is getting the axb5 and Be4 moves for it!
>>
>>Diep nowadays is also completely missing that after the above line
>>white has a won opposite bishop endgame, and it's hard to blame it for that!
>
>
>I wouldn't be so sure. Their "opposite colored bishop" code is much more
>sophisticated than you believe. I had a discussion with Hsu at one of the
>last ACM events, where he noted that CB's "opposite bishop" score was wrong
>in a specific case. He said he had the same problem and in chatting with a
>GM about the problem, the GM pointed out that there were _plenty_ of exceptions
>where opposite B's don't draw. He showed me some test cases where we got the
>wrong answer and ran them against DT where it got them right.
>
>Their code isn't just "opposite B's == bad". It is _much_ more.
Mine is also much more as it might seemingly be, but in this position
both Deep Blues move as well as the score i get after the big fail low
(0.10) is explained easily by the high penalty for the pattern.
This is an example of an opposite bishop endgame
which is completely won for white.
I admit it's real hard to see that white can go to g7 with his king,
so it's hard to blame the eval, but nevertheless this high penalty
explains why it plays Be4 very clearly.
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.