Author: John Merlino
Date: 14:06:37 04/19/01
Go up one level in this thread
On April 19, 2001 at 16:45:16, Scott Woods wrote: >Hi John, > >My tuppence worth.... So many points raised in just 1 post..... sorry. > >When I play CM8K default settings(64Mg Hash) against the Ubtzinger?? settings ( >Also 64Mg hash ) with their respective opening books - the games always seem >pretty even to me. (This is on 2 PC's by the way ) > >QUESTION: Are you are playing CM against itself with the different settings too? >======== Yes. All I am testing is custom books against the default book on one machine. Apart from changing the hash table size to 8MB, the personalities are identical to the default CM. One personality uses CM8000.OBK, and the other one uses the custom book. The test is done on a single PIII-600 machine. > >However against another PC package;Fritz 6 for example, the Ubtzinger?? settings >seem to always come out of an opening book in a much better position than the >standard CM8K offering. >( In Cm standard it always come out in a position loss ) > "Always" is a bit strong of a word. Do you have any statistics? >There are one or two bits here that need some thought and knowledge ( admittedly >more I have. ) > > >I read on this site that Johan uses a special anti-comp book for his >tournaments. >Well I can believe that. >I'm no chess expert - but even I feel that the standard CM opening book is out >of the dark ages in comparison to the competition. This is not exactly true. The book that Johan uses is not and was never actually INTENDED to be an anti-comp book. According to Johan: "It was originally inteded to confuse human players in the famous Aegon tournament (...1997). It turned out to work pretty well against other programs as these could also get confused in messy positions." > >Then there are the engine parameters. >Previously you have posted that the default settings are the strongest for all >types of play. > >QUESTION: Does this mean that... >========= > >1/ These are the optimal settings for all particular types of game. > But not neccesarily the strongest for Blitz, or 60mins or whatever.... >or >2/ No matter what the time controls - the CM team believe there is nothing more > that can be tweaked in the parameters to make CM stronger in any particular > type of game. > >I strongly suspect the answer is option 1. > You are correct. The settings are intended to be the strongest possible for all time controls AND all possible hardware configurations. Remember that our minimum spec machine is a PII-233 with only 48MB of RAM. >As other posts here have said is that when Johan operates the computer he gets >much stronger results. > >I guess this is because... >a/ He knows what software package he will be up against & he has done his > homework. >b/ He can tweak the opening book and settings to play a style of chess > that is capable of bettering that particular opponent. > >c/ My experience is that "The King" seems susceptible to low performance on >slower hardware more than it's opponents. > ( as it appears not to use tablebases - it needs to think about the endgame > a lot more ) > >QUESTION: Can you clarify? > Certainly. First of all, sadly, the new SSDF results are still being looked into. It is quite a surprise that the results are similar to when The King was getting only half of the CPU. Johan is currently looking at the more recent games to determine if a problem still exists in the SSDF's testing (or in the engine itself). Therefore, the results "when Johan operates the computer" should THEORETICALLY be similar to when anybody else operates the computer with similar hardware and time controls. I do know that Johan does quite a bit of analysis and tweaking, but I also know that he uses the default personality settings (apart from modifying the hash table size, depending on the time control). I do not know if Johan actually adjusts the settings between rounds in a tournament, though.... Without access to tablebases, The King is at a disadvantage in endgames compared to programs that do use them. There is no doubt of this. The next version of Chessmaster is almost certain to have some kind of endgame database support. > >Now here are some marketing opportunities for you... > >Invite your customers to.... >1/ A CM8K competition to find the best performing engine setting/ opening book That's always a decent thought. Coordinating it, though, would be tricky. What happens if we get 1000 (or more) entrants (which is likely)? We don't exactly have the manpower/hardware to run that large of a tournament, even if it is single-elimation. > >2/ Identify what they want in the next CM release. > ( Erm, tablebases, learning etc ) That's why I monitor this board.... ;-) > >3/ Why not use some customers as EXTERNAL beta testers > If there is a problem with the software - they will find it as the > previous posts on this site have shown - and most likely well before > the final release date..... hey no patches! > > I'm sure you can round up some bodies for free. > This, of course, would be very useful. However, the powers-that-be do not allow us to use external beta testers. The only time we have been able to release Beta versions of the software to the public was when we did the USCF rating calibration test. Other than that, all testing is done in-house with our own testers. > >Thanks in advance >Scott > No problem. Always glad to help, 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.