Author: Robert Hyatt
Date: 10:19:24 10/18/00
Go up one level in this thread
On October 18, 2000 at 10:43:41, Uri Blass wrote: >On October 18, 2000 at 10:32:03, Robert Hyatt wrote: > >>On October 18, 2000 at 01:20:51, Christophe Theron wrote: >> >>>On October 17, 2000 at 17:29:04, Robert Hyatt wrote: >>> >>>>On October 17, 2000 at 13:09:52, Christophe Theron wrote: >>>> >>>>>On October 16, 2000 at 23:21:23, Robert Hyatt wrote: >>>>> >>>>>>On October 16, 2000 at 21:57:28, Sune Larsson wrote: >>>>>> >>>>>>>On October 16, 2000 at 21:10:04, Robert Hyatt wrote: >>>>>>> >>>>>>>>On October 16, 2000 at 19:51:41, Chessfun wrote: >>>>>>>> >>>>>>>>>On October 15, 2000 at 21:35:33, Robert Hyatt wrote: >>>>>>>>> >>>>>>>>>>I have to disagree. It is not hard to tune my (or any other) program to >>>>>>>>>>play this move. If you watch gambit tiger play, it has some _outrageous_ >>>>>>>>>>scores. In a game on ICC the other night, Crafty was at -.2, gambit tiger >>>>>>>>>>was at +3.2... >>>>>>>>> >>>>>>>>>Dr. I am interested in seeing this game. >>>>>>>>>I assume it was against subtleone as I currently see 9 in >>>>>>>>>it's history. Can you advise which game it was. >>>>>>>>> >>>>>>>>>Thanks. >>>>>>>> >>>>>>>> >>>>>>>>I honestly don't know. Albert can answer as we were chatting during the >>>>>>>>games... he was kibitzing tiger scores, crafty was kibitzing its own scores. >>>>>>>> >>>>>>>>About all I can do to help is to say crafty was white, it was an opposite >>>>>>>>castling game (crafty on queenside, tiger on the kingside). I came in right >>>>>>>>after the opening so I didn't notice what it was. And due to distractions I >>>>>>>>don't know how it ended. I simply remembered +3.2 from tiger, and -.2 for >>>>>>>>Crafty... until finally Albert said something like "+.5 here now, it seems >>>>>>>>that the attack is over..." >>>>>>>> >>>>>>>>Bob >>>>>>> >>>>>>> >>>>>>>Maybe it was the following game: >>>>>>> >>>>>>> >>>>>>> >>>>>>>[Event "ICC u 5 3 2000.11.02"] >>>>>>>[Site "Internet Chess Club"] >>>>>>>[Date "2000.11.02"] >>>>>>>[Round "-"] >>>>>>>[White "crafty"] >>>>>>>[Black "SubtleOne"] >>>>>>>[Result "0-1"] >>>>>>>[WhiteElo "2935"] >>>>>>>[BlackElo "2912"] >>>>>>> >>>>>>>1. e4 c5 2. Nf3 Nc6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 d6 >>>>>>>6. Bg5 e6 7. Qd2 Be7 8. O-O-O O-O 9. Nb3 Qb6 10. f3 Rd8 >>>>>>>11. Be3 Qc7 12. Qf2 d5 13. exd5 Nxd5 14. Nxd5 Rxd5 15. Rxd5 exd5 >>>>>>>16. g4 Bf6 17. Kb1 Be6 18. c3 Rc8 19. Bxa7 Nxa7 20. Qxa7 Qf4 >>>>>>>21. Bg2 b5 22. Rc1 Qxh2 23. Qf2 Qf4 24. Qd2 Qxd2 25. Nxd2 d4 >>>>>>>26. Ne4 b4 27. a3 bxc3 28. bxc3 Be7 29. cxd4 Ba2+ 30. Kb2 Bxa3+ >>>>>>>31. Kxa3 Rxc1 32. Kxa2 Rc2+ 33. Kb3 Rxg2 34. d5 f5 35. gxf5 Kf7 >>>>>>>36. d6 h5 37. Nc5 h4 38. Ne6 Ke8 39. Nf4 Rf2 40. Kc4 Rxf3 >>>>>>>41. Ne2 h3 42. Nd4 Rf2 43. Kd3 h2 44. Ke3 Rg2 45. Kd3 h1=Q >>>>>>>46. f6 Qd1+ 47. Kc4 Rg4 48. f7+ Kxf7 49. Kb5 Qxd4 50. d7 Qd5+ >>>>>>>51. Ka6 Qc6+ 52. Ka7 0-1 >>>>>>> >>>>>>>I saw 5 games played 16/10 between Crafty and Gambit Tiger (SubtleOne) >>>>>>>Result: 1-4 >>>>>>> >>>>>>>Sune >>>>>> >>>>>>I don't believe that was it. I am almost certain it was a draw, >>>>> >>>>> >>>>> >>>>> >>>>>Bob, getting the draw after trying a big speculative attack is not exactly what >>>>>I would call a failure... >>>>> >>>>>There was no winner, but at least somebody tried something... >>>> >>>> >>>>Of course it isn't... but that was an example only. I have seen games where >>>>the big score turned into a lost endgame. I only noticed that one as our >>>>scores were over +3 _different_. >>> >>> >>> >>>But I still don't understand. It is also possible to think you are winning a >>>knight, take it, get a +3 score, then lose the game a few moves later for any >>>reason (checkmate, trapped queen, unstoppable passed pawn pair, whatever). >>> >> >>The question would have to be "how often does that happen?" I don't see it >>often enough to be able to cite a single game. The question has to be, which >>is more prone to errors. I would say sacrificing pieces for mythical advantages >>is more dangerous than the case where you think you win a piece but don't, >>because the former will happen far more often, IMHO of course. > > >I believe that the choice to use gambittiger and not the default tiger suggest >that the change in gambittiger is not wrong inspite of the fact that christophe >did not finish to tune the evaluation. > >Uri I would hardly say his approach is _wrong_. I only said it is _suspect_ and found a couple of cases where the evaluation itself was probably wrong, even if the overall approach works well. The problem here is "full width". If you have a fast, deep searcher. And you have a broken piece of evaluation, such a big tree will let the program search into many positions where the damanged code can screw up and influence the root score in a bad way. Something that is 90% right won't cut it. Because that 10% will be used so much in today's huge trees, it will definitely influence overall results.
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.