Author: blass uri
Date: 04:13:13 08/04/98
Go up one level in this thread
On August 03, 1998 at 23:33:18, Ed Schröder wrote: >>I mean this: > >>" >>Because many people (besides playing games and analysis) also >>use a chess program for solving difficult chess positions Rebel10's >>chess engine has been improved on several points resulting in >>incredible improved solution times. > >>Rebel10 when analysing a position (automatically) will activate >>a special search algorithm that is fully tactical based and will >>find key moves up to 500-1000% faster. The average performance of >>finding key moves faster is estimated at 250-300%. " > >I see what you mean. The above quote has nothing to do with >playing normal games. Also there is no reference to playing >strength as you indicated. > >I refer to what Fernando posted and also to one of the latest >CCC opinion polls. Through the years people started to use chess >program in another way. In the early days people (mostly) >played games against chess programs. These days 99% of people >lose all the time so the fun goes away. > >So people looked for other ways to use chess programs. Based >on all the input I have received through the years (emails / >letters) I would say we have the following situation (for the >use of the engine only)... > >- Play normal games (strongest settings) 10% >- Play handicapped games 20% >- Use chess program for analysis 60% >- Use chess programs for COMP-COMP games 10% > >All estimated of course, just my personal opinion. > >In this respect it makes a lot of sense to improve the "analysis" >part of Rebel. The above new feature of the engine is just an >extra. People want faster solution times for tactical positions >so they get what they want. Also a new fashion is automatic EPD >analysis as EPD is becoming more and more popular. So to make >(keep) your program attractive you write new EPD features. And >so on. > >- Ed - I agree it makes a lot of sense to improve the "analysis" part of Rebel. 1 option of analysis is to give the program to play against itself the main lines for some moves and create a tree of sensible moves. This option in fritz5 is called correspondence analysis. The correspondence analysis of fritz5 is silly because fritz5 does not use the alpha beta algoritam in the tree created by the games it plays and can analyze lines there is no reason to analyze. Fritz5 shows evaluation function only after the leaves of the tree and I prefer to see evaluation function after every move in the tree it creates. Another problem is that I cannot tell fritz5 to use more time in the games it plays against itself in moves near the initial position it analyzes. Uri I want to know if you are going to do a correspondence analysis better than fritz5's Uri > > >>Jouni
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.