Author: blass uri
Date: 20:34:30 03/16/00
Go up one level in this thread
On March 16, 2000 at 20:10:46, Dann Corbit wrote: >On March 16, 2000 at 17:13:47, blass uri wrote: >>On March 16, 2000 at 14:02:28, Robin Smith wrote: >>>On March 15, 2000 at 15:08:36, blass uri wrote: >>>>On March 15, 2000 at 14:17:59, Robin Smith wrote: >>>><snippped> >>>>>In the 13th US correspondence chess championship finals I played a move that my >>>>>computer NEVER finds. Do any programs find 17.... Qd5! >>>>> >>>>>This move prevents white from playing Be4 (which would put presure on b7 and >>>>>support a possible d pawn push), it frees d8 for blacks rook and it keeps the >>>>>queen eyeing d5, restraining whites isolated d-pawn. I believe it is the >>>>>strongest move in the position, but when I reviewed the game with my computer it >>>>>never finds it even after a very long "think" because of the exchange sac >>>>>involved. >>>> >>>>Did you try to generate a tree with the computer to prove that 17...Qd5 is >>>>really the best move? >>> >>>I don't know how to "prove" that a positional move is best. I do know that it >>>seems programs reject it out of hand because of a faulty "refutation" ... the >>>game continuation. >> >> >>Computers always prove that a move is best by a tree. >>one of the problem of programs is that they do not know to do the right >>extensions. > >But "BEST" is only valid for the exact depth of the tree. If 12 ply has been >analyzed, the tree can change at ply 13. Indeed, there is approximately 15% >chance it will do so from an arbitrary position according to the research by >Hyatt and Heinz. I agree that what I defined as a proof by a tree is not sure but I think that using this method is better for long time control relative to giving the computer hours to analyze only the root position because humans know better which lines to extend relative to computers. > >>I can generate a tree and prove by telling programs to give me more than one >>option that there is no good move out of the tree. >> >>I can calculate the best move based on the evaluations after search of the >>leaves of this tree. >> >>If the best move based on these evaluations is Qd5 I see it as a proving tree >>that Qd5 is best. > >I don't believe any computer analysis is sure unless it ends in checkmate for >one part or the other or in stalemate. Even then, it is not sure if NULL move >is enabled and if not then how will it get analyzed deeply enough? > >> >>>>It is clear that after 19.Nb6 white is losing but the question is if white has >>>>no good alternative before(for example 18.f3) >>> >>>18.f3 is maybe better than what white played, but white's exposed king, weak >>>d-pawn and blacks 2 bishops are very strong strong advantages. Either 18.f3 Rd8 >>>or 18.f3 Qg5 should give black a winning position. >> >>You may be right about f3 >> >>I did not analyze f3 for a long time because it was not my game. >>I got with Junior5.9 a better score for black by 17...g6 relative to 17...Qd5 >>18.f3 and this was the reason that I suggested this move but I see that the >>Junior5.9 intends to play the same wrong idea of Qa4 Be7 Nb6 after 18.f3 Rd8 >> >>The point is that I cannot be sure that black is winning without analyzing a lot >>of lines with the computer. > >In the final position it is surely winning after the choice of Qd5 has been >followed through, and I think he is right that the lines are fairly well forced. I think that Nb6 is not forced and chessmaster6000(ss=10) rejected Nb6 after some hours and found Ne3. Qa4 is also not forced and the question is if the evaluation after what computers consider as the second best move instead of Qa4 or the evaluation after Ne3 instead of Nb6 is better for black relative to the evaluation of the root position when computers want to play moves that are not Qd5. 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.