Author: Uri Blass
Date: 17:19:33 03/14/04
Go up one level in this thread
On March 14, 2004 at 19:18:37, Mike S. wrote: >On March 14, 2004 at 12:54:04, Uri Blass wrote: > >>Is CSS rating list played with pondering and at 120/40 time control? > >No, 10m+10s without pondering. Thanks It means that we can learn nothing from CSS rating about ssdf rating. A program may have a bug in pondering that make it significantly weaker with ponder on. > >What is more representative for the usual practise? I'd say shorter time >controls than 40/2h, because I guess only correspondence players - like you, >though :-)) - use more than 1 minute per position in games or analysis often >(and maybe some grandmasters and their seconds, but they aren't relevant SSDF >audience). Also, most computerchess users will have single cpu computers (or two >which are not identical) and therefore no ponder tests are much more relevant >IMO. > >There is no definition of conditions for a "true" strength. > >Another example: The WBEC Ridderkerk competitions are played with 40/40m, ponder >on (I think on a dual; I'm not sure at the moment). Very good condititions IMO, >a kind of compromise. > >Anyway, if SSDF would decide to shorten the time controls to 40/40m or even >60/30m or the like, I'd appreciate it. I prefer that the ssdf continue to test at 120/40 There are enough testers who test blitz games and we need tests also at 120/40 time control. The main problem currently is, that >engines don't play equal numbers of games against the same opponents for the >SSDF ratings, unlike in the CSS ratings which are done like that (except the >head-to-head match of course, when two engines are compared). These are very >different concepts (if the SSDF method to chose the opponents has any concept; >it seems chaotic to me). IMO the CSS tests are much more on an equal basis. > >It must raise doubts when the average opponent's strength is so different, like >it occured with the Ruffian 2 test recently, for the SSDF ratings. I do not think that it is the problem. Ruffian2 simply performed poorly with ponder on also in other tournaments(see WBEC) 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.