Author: José Carlos
Date: 05:36:02 11/15/01
Go up one level in this thread
On November 15, 2001 at 07:16:14, Uri Blass wrote: >On November 15, 2001 at 06:51:23, José Carlos wrote: > >>On November 15, 2001 at 03:35:23, Dieter Buerssner wrote: >> >>>The end position in this game is >>> >>>[D] 8/8/6p1/6Q1/2K4p/6kP/5p2/8 b - - 0 55 >>> >>>How does your engine evaluate this pos? >>> >>>[Event "?"] >>>[Site "Beersheba ,CBM 09"] >>>[Date "1988.??.??"] >>>[Round "?"] >>>[White "Kortchnoi, V."] >>>[Black "Rechlis, G."] >>>[WhiteElo "2640"] >>>[BlackElo "2475"] >>>[ECO "A15"] >>>[Result "1/2-1/2"] >>> >>>1. Nf3 Nf6 2. c4 g6 3. Nc3 d5 4. cxd5 Nxd5 5. g3 Bg7 6. Bg2 >>>O-O 7. O-O c5 8. Qa4 Nc6 9. Qc4 Nxc3 10. bxc3 b6 11. Qh4 >>>Bb7 12. Rb1 e6 13. c4 Qxh4 14. Nxh4 Rab8 15. Nf3 Nb4 >>>16. Bb2 Rfd8 17. Bxg7 Kxg7 18. Rb2 Rd6 19. d3 Re8 20. a3 >>>Na6 21. Ra1 e5 22. Nd2 Bxg2 23. Kxg2 f5 24. a4 Nb4 25. a5 >>>Re7 26. f3 h5 27. h3 Kf6 28. axb6 axb6 29. Ra8 Rdd7 30. Rb1 >>>Ra7 31. Ra1 Nc6 32. Nb1 Rxa1 33. Rxa1 Ra7 34. Rxa7 Nxa7 >>>35. Nc3 Ke6 36. Nd5 b5 37. Nc7+ Kd7 38. Nxb5 Nxb5 39. cxb5 >>>Kc7 40. f4 exf4 41. gxf4 h4 42. Kf2 Kb6 43. Ke3 Kxb5 >>>44. Kd2 Kb4 45. Kc2 c4 46. Kb2 cxd3 47. exd3 Kc5 48. Kc3 >>>Kd5 49. d4 Ke4 50. Kc4 Kxf4 51. d5 Kg3 52. d6 f4 53. d7 f3 >>>54. d8=Q f2 55. Qg5+ 1/2-1/2 >> >> This is the kind of position programs can chose the right moves, but don't >>understand at all. My program says +7.xx for white, but it defends correctly >>with black. >> I guess this position won't be solved by any program in a long time because it >>requires reasoning. You can't write code for positions like this easily, because >>an extra pawn here or there makes a difference. So you have to 'think'... the >>magic word. > > >I am not sure that this position will not be solved by search > >I suspect that it is also possible to see forward enough to see a forced >repetition or tablebase position so it is possible that tablebases can help. > >Based on the post that I read they did not help but maybe the poster did not >give the programs enough time. > > The word that puts distance between GM's and programs. GM's can, >>given enough time, understand _any_ position. > >I doubt it > > And if they misunderstand one and >>lose, the next game he'll have learned the lesson. Computers can't do such a >>thing. > >Humans may learn not to lose by similiar way but it is possible that the >position included a very complicated tactics and the GM's learned to avoid one >mistake only to do another mistake. > >Uri I didn't express myself clear. I was talking about positions that require understanding, not any position including tactical ones. What I meant is: A GM has a quiet (it's not solved by a tactical shot) position on the board. He _thinks_ about it. He sees some details that make the pawn down compesated. He says 'I need to attack the weak square at g6 because otherwise my opponent will simplify and I'll be a pawn down in the endgame. So I gotta move my pieces to the kingside. I don't care about my pawn in a2; if I don't attack, I'll lose the endgame anyway. So I'll move my pieces towards g6 and he can take the a2 pawn. My pieces are not very well coordinated so I'd better not open the position yet... A program can't think like that. A program would defend the a2 pawn. Or if it has the knowledge that being down in material it has to attack, it'll try to open the position and anttack, etc... Every decision the program makes is very simplistic compared to what a human can think of. And if the GM is wrong, he'll later analyze the game and look for the misconception. And game after game, he'll become stronger. No program can go from being a 1500 player to 2500 without being reprogrammed. Humans are not reprogrammed. Humans _understant_. Humans _learn_. That's my point. José C.
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.