Author: pavel
Date: 14:54:57 05/09/05
Go up one level in this thread
On May 09, 2005 at 08:05:48, Christos Gitsis wrote: >On May 08, 2005 at 18:00:01, pavel wrote: > >>All games (so far) >>-------------------------------------------------------------------------------- >>Program Elo + - Games Score Av.Op. Draws >> >> 1 Aristarch 4.50 : 2510 6 6 9463 52.8 % 2490 24.7 % >> 2 Yace Paderborn : 2490 6 6 9463 47.2 % 2510 24.7 % > >Hello, pavel > >I always thought that Yace Paderborn would be more than 100 elo weaker than >Aristarch 4.50 especially since it is an old version. >Your results show that the two programs are very close at strength. >This is encouraging for all Yace fans. :) Chris, I think Aristarch is atleast 40-50 elo stronger than YacePB in tournament time control, Because Yace will never play enough games to take advantage of it's positional learning in such a long time control. I think what this test result shows is that positional learning is far more important than most people give it credit for. I wish there was some kind of client/server system to share positional learning by all Yace users. Like Antivirus, Yace positional learning will get updated once a day. Let's say there are 1000 Yace users. Each user will automatically upload it's positional learning (only new learned positions) to a central server, which will than filter out the duplicates (if any) and than send it out to all users as update, once a day. The same could be done with book learning. I think in this way the program strength can be increased by atleast 200-300 points. Ofcourse the current Yace limit of 100,000 learned positions has to be increased. :) Maybe all this doesn't make any sense, but it's not really an impossible scenario. IMO. Cheers, pavs
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.