Author: Heinz van Kempen
Date: 07:50:16 01/17/06
Go up one level in this thread
On January 17, 2006 at 09:03:00, Vasik Rajlich wrote: >On January 16, 2006 at 13:21:33, Heinz van Kempen wrote: > >>Hi all , >> >>CEGT downloads and rating lists 40/40 (2 Ghz) are updated. Hopefully everyone >>will find a list telling most. >> >>http://www.husvankempen.de/nunn/downloads.htm >> >>http://www.husvankempen.de/nunn/rating.htm >> >>For the 64-bit versions I kept those giving big differences still. There is >>still an additional database where all is separated in case that there will be >>changes in the future rating. >> >>Best Regards >>Heinz >> >>http://www.husvankempen.de/nunn/ > >Heinz, > >a quick glance at the table suggests something that other testers have also >seen: the optimistic setting does better against weaker opponents. > >I am curious, what is your position about having an engine setting through which >the tester indicates the playing level of the upcoming opponent? > >Also, for your information, Beta 9 was an intermediate version between Beta 7 >(which had the same playing strength as Beta 1) and Beta 10. I expect that >whatever difference exists in Beta 9 is magnified in Beta 10. Also, neither >version will be able to keep the pace of your current results. > >Vas Hi Vas, first of all a clarification, because some thought that opt stands for Opteron. opt stands for optimistic setting and the games are played on Athlon64x2 4200+ and Athlon64 3500+ with CEGT 40/40 (2Ghz) adapted time control what is 40/18 repeated on this hardware due to Crafty benchmark. Still too early to tell if the optimistic settings work very well only against "weaker" opponents (those were all from Master Class, so currently the best we have), because I am still having ELO performance higher that 3000 against some of the strongest currently. How do this optimistic settings work? Is their a bigger tendency to do sacrifices or is the evaluation generally higher than with default, so that we have some sort of contempt value? Anyway I took this beta 9 for Master Class because it was the latest available when I started. It will win the second tournament by a scary margin. So now I am curious if the rating can hold with more games what would be unbelievable, but for the moment it is still unclear where the limits are, if there are any at all. No concentrated tests can be currently run for engines with 64-bit, because we are still lacking testers with 64-bit OS and four of those machines are now used for the leagues. Rybka 1.01 beta 9 64-bit will receive 300 games before testing the next version, what will probably be the release 1.2 then. Well, people were criticized here because of posting too much about Rybka. I don´t care. Just watching the games it is hard to hold back enthusiasm. Sacrifices for positional advantages, slowly suffocating the opponents by just conquering dominating positions for the pieces, extremely good evaluation of passed pawns and bishop pair, active king in the endgame. We should also find the time to comment some of the games and not only the results. On the other hand don´t let us forget other engines improving and not noticed so much. I did not comment so far, but for example the new Spike 1.1 beta is having great results in CEGT, so more exciting stuff to come. Best Regards Heinz
This page took 0.01 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.