Author: Uri Blass
Date: 01:32:21 03/24/02
Go up one level in this thread
On March 24, 2002 at 04:04:15, Thorsten Czub wrote: >On March 23, 2002 at 18:31:32, James T. Walker wrote: >>Why? Because Chris himself told me this. Who knows more about CSTal than >>Chris? Certainly not you. > >chris himself :-)) > >first: i don't think chris knows more about cstal than i do :-)) >second: of course he began a complete rewrite. at was at the time >when version 2.5 was there. > >i have no idea what purpl made out of it. > >what i wanted to say with WHY was: there is no need for a rewrite. >the old and rusty algorithms from 1996 or 1997 are good enough for fritz7. I disagree I did not test tal dos so I cannot be sure about it but even if the old algorithms are good enough for Fritz7 then it does not mean that they will be good enough for tiger15 or for fritz8. Even if the old algorithm are good enough to get 60% against the top programs(and I know no evidence that they are good enough even to score 50%) then it does not mean that there is no need for a rewrite in order to get 70% against them. > >the difference between NPS between fritz7 and cstal1 (dos) is >between 17-20 times. Nodes mean nothing for me. Program can search more or less nodes if you change the definition of nodes. > > > >>>33.000 NPS is enough for fritz. >>>Because Fritz is artificial stupidity. > >>And CSTal is artificial snail. It is interesting and fun to play and sometimes >>plays brilliantly but against other computers like Fritz it gets out searched >>and loses more often than not. > >how do you know that ? > > >>I still think the first game loss was due to a >>poor book. In the second game it had a better opening and played to a draw. >>Two games tell you nothing. > >who tells you that i do only have 2 games ? > >>Don't know. I haven't tested it lately. Used to be 100+ ELO. Now maybe a >>little less, maybe a little more. When I ran it on ICC it was constantly >>100-200 below top commercial programs. This was on a K6-2-350 and later a >>K6-3-450. > >maybe the other programs made a step backwards because they had to cope with >king attacks and new paradigm. >maybe the other programs made no real progress. >ever tried fritz5.32 versus fritz7 ? Fritz7 is better than Fritz5.32 based on all the rating list that I know. Fritz5.32 may score better in one short tournament but I do not know about a serious tournament with 100 games for every program when Fritz5.32 can score better. Fritz7 got second place after the king in a tournament that was published here and it is the version with the long castling bug(7.002) Do you believe that Fritz5.32 could get similiar results? 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.