Author: blass uri
Date: 07:38:52 07/29/98
Go up one level in this thread
On July 29, 1998 at 10:14:09, Mark Young wrote: >On July 29, 1998 at 04:16:08, blass uri wrote: > >>I began to do a nunn match with 30 minutes per game+30 seconds per move >>and Fritz5 has better results than with 10 minutes per game+10 seconds per >>move(in that match after 10 games the result was 7.5:2.5 for Junior5 and now the >>result is 5.5:4.5 for fritz5). >>I gave both programs 59968 Kbytes hash tables >>in the 30 minutes+30 seconds match when I gave both programs almost 12000 Kbytes >>hash tables in the 10 minutes+10 seconds match. >>. >> >>Fritz5 has a weakness in seeing the position >> >>It is the reason it lost to stobor in the World micro computer championship. >>It traded queens and only after it did it it saw stobor's position is better and >>it is not a question of time(even if I give fritz5 30 minutes it will do the >>same mistake). >> >>Something like this cannot happen to Junior. >>This is the reason I expected Junior to earn more from time than fritz >>but it did not happen. >> >>Uri > >I have found the same thing with fritz 5. When people say it plays weak moves I >have to wonder if they were playing it at 5 sec. a move. I found that fritz 5 >gains rapidly in strength as it has more time. Much more so then other programs. >Except for maybe CM 5500 or 5000. The weak move it played against stobor in the WMCCC was not at 5 sec a move and the reason for the mistake is that fritz5's evaluation function is sensitive to the root position. This is a bug in fritz5 because it can be not sensitive to the root position in the nodes that are near the root. I think the reason Fritz5 gains more than other programs from time is that it has other advantages (for example better positional understanding) Did you do tests with fritz5 against other top programs(except Junior)?. 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.