Author: Uri Blass
Date: 11:34:03 01/02/01
Go up one level in this thread
On January 02, 2001 at 14:01:27, Bruce Moreland wrote: >On January 02, 2001 at 06:08:12, Uri Blass wrote: > >>On January 02, 2001 at 04:20:12, Bruce Moreland wrote: >> >>>On January 01, 2001 at 23:54:02, Pete Galati wrote: >>> >>>>Ok, I gave Comet 72 hours to look at this position, no changes for over a day, >>>>it never reached SD 17. It was doing approx 60mb hashtables in Windows ME's >>>>Dos. Comet seems quite sure that Nf3 is the correct move. Personally, I would >>>>have worked on stacking the Queen's Rook over the Queen. >>>> >>>>There was 2 games between Spassky and Petrocian (spelling?) in Moscow in 1969 >>>>with the same position. The one that's 28 moves long is interesting because in >>>>that one Spassky got to take him apart for a few more moves and then things made >>>>more sense to me. >>>> >>>>Was e5 _really_ the best move? >>> >>>e5 is shattering. I have no problem believing that it is the best move. >>> >>>bruce >> >>I agree that e5 is winning but I cannot see +5 advanatge for white even after >>some analyis with chess programs and the best that I can see is scores that are >>close to +3. > >I ran it for several days on a quad processor, and I think that my program is >known for finding tactical combinations. I believe that your program saw the tactics and I guess that it is better than most programs in the e5 position. > >>It will be interesting to get a proving tree to the +5(I mean a tree when >>programs can see in every leaf of the tree in a few minutes +5 evaluation for >>white or at least +4). > >I don't need to do that, the alpha-beta algorithm should do that for me. You are right that the alpha beta is enough to prove it but I think that these proving tree may help to learn which lines to extend. > >>The score for the other nolot position is more convincing because black must >>follow the main line. <snipped> >The score for that is perfectly convincing. If the line it produced was +5, the >main line must be as bad or worse. It is possible that the program will fail >low down to something a little lower than +5, but this is not very common with a >score like +5. That's almost always a crush. I believe that the score is correct. When I say that the score is not convincing I mean to say that I do not understand why the score is correct. I admit that I tried to do it with programs only for a few hours and Ferret used more time and better hardware but it is known that computers are using most of their time for analyzing illogical lines. I also knew the main line of the solution when Ferret did not know it before it started to search. This is the reason that I hoped to find a proof for the +5 in relatively a short time. 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.