Author: Uri Blass
Date: 07:03:05 12/21/02
Go up one level in this thread
On December 21, 2002 at 09:36:01, Enrique Irazoqui wrote: >On December 21, 2002 at 09:19:16, Uri Blass wrote: > >>On December 21, 2002 at 08:58:40, Enrique Irazoqui wrote: >> >>>On December 21, 2002 at 06:48:02, Torstein Hall wrote: >>> >>>>After trying F8 on the playchess server and in engine tournaments I find that F8 >>>>is not so good in Blitz games. >>> >>>I have the same impression. I also played an engine-engine match on an AMD 2200+ >>>at 5+3, 96 MB hash each, own book each, and DF7 beat F8 71-35. >>> >>>> My Playchess rating fell about 100 points when I >>>>changed to F8 from F7. And in my Engine tournament it is trailing Fritz 7, Fritz >>>>5.32 and Schredder 6. >>>>Fritz 8 may be better in longer games >>> >>>I doubt it very much. I think DF7 is better here too. >>> >>>Enrique >> >>I am interested to know if you tested Deep fritz7 and Fritz8 in your secret test >>suite and if you found that Deep fritz7 is better > >DF7 scored better than F8, but not by much, only 11 points. In that case it suggests that the difference between DF7 and F8 is smaller in long time control. > >>(another question is if your >>tests suggest smaller difference at longer time control and if Fritz8 get closer >>to Deep Fritz7 at longer time control). > >I only use my test at slow time controls. Yes but you can also calculate results at faster time control by calculating the scores of all the programs when you multiply the times by 10(it means that there is no difference between not solving and solving in more than 1/10 of the time for the score of the test for blitz). > >>It may be also interesting to know results of knightdreamer in your secret test >>suite(the author of knoghtdreamer claims that knightdreamer is good in tactics). > >I don't have it. Knightdreamer is a free winboard engine and You can download it in the following address: http://www3.tripnet.se/~owemelin/johan/KnightDreamer.html Based on it's public result in the gcp test suite it seems to be better than part of the commercial programs in the gcp test suite. It may be interesting to know if it is also the case in your test suite. > >>It may be also interesting to know result of my latest movei in it but you may >>wait because I plan to improve the tactical strength of movei in the future and >>there are other interesting programs like Ruffian or List to test(unless you >>already tested them). >> >>Movei is alreay faster than Crafty in some tactical positions but I guess that >>Crafty is still better than movei in your tactical test suite(It seems not to be >>the case in the GCP test suite based on part of the results of that test suite >>but movei is training on this test suite so it is not fair to compare on test >>suite when I know the positions). >> >>Note that I do not say that movei is tuned for the gcp test suite because I only >>add changes that I believe that are good for games but part of my changes are >>results of analysis of failures in the gcp test suite so the changes that I do >>are not random changes to make the program better. >> >>Uri > >I'll be glad to test Movei whenever you wish. Let me know. > >Enrique Thanks I will send you the latest version in a few days or few weeks. Uri
This page took 1.82 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.