Author: Uri Blass
Date: 17:17:19 09/09/01
Go up one level in this thread
On September 09, 2001 at 19:54:48, Gian-Carlo Pascutto wrote: >On September 09, 2001 at 19:37:13, Uri Blass wrote: > >>I did not analyze all these lines but the interesting question is if there is a >>limit of number of plies when you are going to say that it is perfectly >>impossible that Deep thought could see so many plies forward. > >My point is that after c5 black has a _winning_ advantage, revolving >around getting an extra passed pawn (not the win of a piece it seems). > >I don't really care how many plies it is. What matters to me is >that DT saw it. And nobody else does. > >>It seems that you try to claim that Deep thought outsearched the top programs of >>today by more than 30 plies. > >Why not? It outsearched Cray Blitz by 20 plies. Cray Blitz wasn't >exactly a pansy as far as search is concerned. No This is your assumption that it outsearched Cray blitz by 20 plies. It is not my assumption. My assumption that it is impossible and I hope that you agree that if the number of plies is big enough it is also impossible. It is impossible that Deep thought outsearched another program by 200 plies because it is impossible that deep thought searched 200 plies forward in the selective search. > >>Remember that the discussion is not only about the question if Bg5 could save >>the game but on the question if deep thought could see that c5 is winning. > >It did. It had a +2 score for it. No Having a +2 score is not a proof that it saw the win. It may be also a bug in the evaluation. > >>It is clear to me that if there is a win the win is positional and the number of >>plies is so big that deep thought had no chance to see it. > >It is clear to me you are in denial. > >Winning a passed pawn is hardly what I call a positional win. >Nearly all lines I posted are forced. I start with the fact that most of the game moves before Bg5 does not seem to be forced so the depth after Bg5 should be very small. > >There are only a few fancy moves: Bd7/Bg7/c4/Nd3 > >>It is important to give your opponent hours to search for every move >>in the game. >>I do not want to check moves that the computer got by searching for only few >>minutes. > >In that case, it should be trivial for someone as experienced as >you to refute all our analysis right? It may take a long time. I did not say that it is trivial, I usually use computer for hours in my correspondence games. I also check the analysis later but the first thing to do is to give my programs hours. If the first move in the analysis that is different is move 7 then you can multiply 7 by many hours per move and you get few days. 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.