Author: John Merlino
Date: 09:44:19 03/19/04
Go up one level in this thread
On March 18, 2004 at 02:16:07, Chessfun wrote: >On March 18, 2004 at 02:03:15, Johan de Koning wrote: > >>On March 17, 2004 at 12:35:08, John Merlino wrote: >> >>>On March 17, 2004 at 06:58:15, Harald Faber wrote: >>> >>>>On March 17, 2004 at 04:21:06, Graham Banks wrote: >>>> >>>>>As Kurt and other CM devotees are aware, I did extensive work on devising CM8000 >>>>>settings, many of which are on Surak's website (my most successful being the >>>>>early CM Titan and the later CM Legend settings). >>>>>Inspired again by the experimentation of others in getting better results than >>>>>the default settings for CM9000, I will unveil CM9000 (The Judge) in my upcoming >>>>>Super Tournament II. >>>>>I'm looking forward to seeing how CM9000 performs with these new settings, >>>>>having been a little disappointed with the default settings. >>>> >>>> >>>>Sorry I am only kidding with CM 9000 WCH (World Champion) settings. >>>>Aren't you getting tired creating new settings which seem to be better than the >>>>original ones and/or the thousand different settings posted by other CM >>>>enthusiasts - and finally are NOT? >>> >>>It's really a question of how you decide if these custom settings are "better" >>>or not. Under the parameters of the tests (which typically involve a specific >>>time control on specific hardware), almost invariably the custom settings >>>perform better than the default. I don't think people would be posting their >>>results if they were not noticeably better than the default under the same >>>situation. >>> >>>But, most of the time, these new settings perform no better than (or sometimes >>>even worse) than the default settings on significantly different hardware with >>>significantly different time controls (tournament vs. blitz, for example). So >>>that is why most of these new settings come with caveats that make it clear that >>>the settings are only for certain situations. >>> >>>The default settings are designed to be as good as possible for the hugely >>>divergent hardware and gaming requirements of the hundreds of thousands of >>>Chessmaster users. Obviously, then, there must be some way to make minor (or >>>even significant) improvements if all you are concerned about is maximizing the >>>strength of the engine on YOUR computer and with YOUR favorite time control. >> >>Well spoken! >>BUT ... >> >>... I'd like to add that YOUR opponent is at least as important >>as YOUR computer x YOUR time control. >> >>People on this forum often talk about "strength" and "stronger" >>while they actually mean "performance against programs X and Y". >>It is definitely cool to score against X or Y. And let's face it, >>it is what's keeping the interest in computer chess alive today. >>BUT ... >> >>... The first goal of CM's engine is to be a reliable mentor and >>analysis tool. Speculative settings, no matter how well they are >>tuned against the latest X and Y, are simply off limits. >> >>... Johan > >There are also those tournaments where the default personality wins! >http://chessprogramming.org/cccsearch/ccc.php?art_id=316859 > >Sarah. And notice the very slow hardware, a P2-400, which is actually below the minimum spec of CM9000. Also, since this tournament was played under the CM GUI, I would bet that the engines had pondering on. So the personalities were getting a lot less CPU time than in most other tournaments that have been reported. It is no surprise to me, then, that the default personality won under this situation, when many of the other personalities were designed for a completely different situation. jm
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.