Author: Uri Blass
Date: 23:39:45 07/19/02
Go up one level in this thread
On July 19, 2002 at 19:59:26, Robert Hyatt wrote: >On July 19, 2002 at 13:45:07, Uri Blass wrote: > >>On July 19, 2002 at 13:16:06, Joshua Lee wrote: >> >>>I was wondering that if the DB team estimated a rating or over 300 points higher >>>than the programs at that time on equivalent hardware or equal search, they >>>weren't clear on that. That would mean 2796 on a P200 and Fritz 7 for example >>>wouldn't be anywhere near this on a P200. Has anyone seen anything about their >>>tests against commercial software and does anyone have any opinions about this? >>> >>>Granted that Software has improved and that on Todays Fastest Hardware The Top >>>Programs (provided the SSDF list is that accurate) is on equal footing with Deep >>>Thought and DB Jr. >> >> >>Why do you think that they may be only equal. >>I believe that the opinion of most programmers is that >>the top programs of today are clearly better >>than Deep thought and deep blue JR. >> >>It is known that their search algorithms were bad >>relative to what is known today >>(otherwise you could see people who copy their >>search algorithms with good results when I know that >>the people who tried to copy the way that they use >>singular extensions got bad results in games so >>they do not use it). >> >>It is known that they believed some wrong assumptions >> >>They believed that null move pruning is dangerous at >>their speed when the results say that null move pruning >>is only more productive when the machines are faster. >> >>They believed that the brute force depth is not very important >>at their speed and singular extensions are more important >>and they were again wrong(programs already >>have similiar speed to deep thought and they do not use >>singular extensions in the way that deep thought used them). >> >>Their speed simply misleaded them to wrong assumptions >>that they did not test. >> >>Uri > > >That last sentence is simply wrong. > >On two counts... > >1. They tested _and_ wrote several papers on their search extensions. It >was hardly "not tested". I assume for this discussion that they tested it in games against a version without null move pruning and without singular extensions. Did they test it in games against a version with null move pruning? It is possible that they could get 2 productive changes: 1)Add null move pruning 2)Not use the singular extensions. It is possible that 2 is not productive without 1 and if they did not start by testing 2, they got the wrong conclusions. Did somebody try to test crafty with their algorithm (no null move pruning and singular extensions)? I guess that it is going to be clearly weaker than the Crafty of today at 120/40 time control. 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.