Author: Uri Blass
Date: 15:33:10 04/19/03
Go up one level in this thread
On April 19, 2003 at 17:45:45, Omid David Tabibi wrote: >On April 19, 2003 at 17:22:19, Jorge Pichard wrote: > >>On April 19, 2003 at 16:28:46, Uri Blass wrote: >> >>>On April 19, 2003 at 16:15:33, Jorge Pichard wrote: >>> >>>>Great we know that most of these programs are rated at least 2500, but what is >>>>the lowest rated program that can still solve it. Is it BAM BAm or possibly >>>>Movei or Chess Genius 3? >>> >>>You should define can solve it. >> >>Simply to select Ba3 is sufficient as long as the program understand that it >>will be better off 10 or 11 moves later. >> >>I just let Chess Genius 3 to analyse the position under two/playes analysis, but >>when I told it to print the moves analysis it simply froze> Now I saw that CG3 >>selected the correct sequense of moves after Ba3 but too bad that I could NOT >>manage to post its Analysis. >> >>>How much time? >> >>As long as it take Movei to solve it, but NOT more than 60 Minutes. > >Genesis solves this in 29 minutes on a 733 MHz system, at depth 15. Does it only fail high after 29 minutes or have a full line? Latest movei failed high at depth 14 on A1000(I was writing this post during the analysis so the time could be slightly smaller) depth=14 +0.01 b2a3 Nodes: 173998087 NPS: 214584 Time: 00:13:30.86 depth=14 +0.30 b2a3 Nodes: 201050595 NPS: 213013 Time: 00:15:43.84 I think that previous versions could solve it faster and it is possible that some pruning rules that I believe to be productive in games were counter productive here. I also believe that the problem can be solved by evaluation and it is possible that botvinik did not see until the end when he played the sacrifice. It is easy to see that Ba3 is leading to at least a draw with chances to win after considering the line Ba3 Qxa3 Nh5+ gxh5 Qg5+ Kf8 Qf6+ Kg8 e7 The point is that the black knight is too far away to stop the pawn and the queen can stop the pawn only by perpetual (there is no way that it can get to the 8th line) so I think that the problem can be solved by evaluation because if your evaluation for other lines is a draw and here you see chances for a win then you are going to give the position a positive score by static evaluation. Movei is a poor evaluator so it does not see it by evaluation and needs a deep search. My extensions rules and pruning rules are also not good enough otherwise movei could fail high faster. Note that movei always does not trust fail high scores so practically it is not going to play this move even after 20 minutes. I know that movei can get wrong fail high even when the difference is +0.30 so I prefer not to trust them. Movei tries to solve fail high and will use more time to do it and only in cases that there is a danger of losing on time it is going to play(in case that it already used more than half of the time that it was allowed to use) 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.