Author: Harald Faber
Date: 21:36:57 11/19/01
Go up one level in this thread
On November 19, 2001 at 01:38:25, Ed Schröder wrote: >On November 19, 2001 at 01:04:46, Harald Faber wrote: > >>On November 18, 2001 at 17:41:22, Thorsten Czub wrote: >> >>>On November 18, 2001 at 14:20:58, Harald Faber wrote: >>>>No, it was just 9 rounds, almost anything is possible. >>> >>>nonsense. the explanation must be in the games, or the machines. >>>or the settings the humans set up. >> >>Sure the explanation is in the games. :) >> >>>do we have information about the hardware that was used and about the settings >>>of openings and style ? >> >>Hardware: >>Gandalf 4.32 UCI - AMD Athlon 1200 MHz >>Junior 7.0 - AMD Athlon 1400 MHz >>Hiarcs 7.32 - AMD 1900+ 1600 MHz >>Gambit Tiger - AMD 1800+ 1610 MHz >>Chessmaster 8777 - AMD 1563 MHz >>Shredder 5.0 - AMD 1500 MHz >>Fritz 7.0 - AMD 1500 MHz >>LambChop 10.87 - AMD 1000 MHz >>Rebel Century - AMD XP 1600+ 1333 MHz >>Nimzo 8 - AMD 1000 MHz >> >>Hash size has been around 184-200MB IIRC except Rebel who only had 256MB RAM and >>under Win2000 200MB hash made Rebel much slower so I took only 100MB hash which >>have been filled within 2 minutes. Might have hurt in some endgame. >> >>>>Nimzo IS weak, Rebel played...not well. >>> >>>nonsense. nimzo8 was never weak. the versions before where shit. >>>but 8 is ok, for the first time since nimzo guernica this nimzo version was a >>>real progress. why do you say it is weak when it isn't. >> >>Because it IS weak. Plays weak moves and is the worst endgame player between the >>commercial programs. >> >>>>Slowest machine was A-1000 for Nimzo >>> >>>aha. NIMZO is weak. it seems to me the hardware of nimzo was weaker than the >>>hardware the others used. what else was with nimzo. setting ? opening book? >>>tablebases ? hash ? >> >>There have been other PCs which were not significant faster, e.g. the 1200 for >>Gandalf. Also Rebel with XP1600+ is only 1333MHz in real. >>Nimzo used the default settings like many other programs. Only addition: >>TablkeBase depth set to 10 (default is 6) to get Nimzo faster to the TBs before >>he loses by bad endgame. Opening book default, complete TBs and around 180-200MB >>hash. >> >>>>Some used own and/or "tuned" opening books, >>> >>>aha. nimzo is weak !! now we come closer to the reasons. >> >>Yes, the reason is the engine. >> >>>> some used the original ones. >>> >>>some did, some not. WHO ? >> >>I am not sure. >>GT2 aggressive of course played with a special book by Jeroen Noomen like Rebel >>did. Junior 7 had some tuned book, edited and modified by Ludwig Bürgin >>(operator of Junior7) in many hours of work. And the last tuned opening bbok has >>been Chessmaster. Longin Bauer had some special books, one for each opponent and >>for each colour. All others have played with the original engine books AFAIK. >> >>>>Settings have been default except Chessmaster which have been by Stefan Kleinert >>> >>>despite opening books you mean. >> >>Special opening books for CM, read above. >> >>>>(CM 8777). Shredder 5 played the last few rounds with "normal" (default is >>>>aggressive). The others played default. Time control in most cases has been >>>>40/120+60. >>> >>>hm.- >>> >>>thank you harald. >> >>After the first three rounds=three losses where I have played 40/120 and changed >>to g/65 after 40 moves I changed to g/180 which gave better results (the loss >>vs. Tiger is normal, best result Rebel can expect vs. Tiger is draw) and didn't >>lose a game since that change. One small reason is: After the change to g/65, >>Rebel moved MUCH too fast, after 30-60sec each move. > >What you describe is an old bug in the time control, it has been there >since the 80's and I everytime forget to fix it. Sorry, I didn't know this. >When you change to "game in xx:yy" in the middle of a game the program >actually resets the time control to move 1 as if it is a new game. The >effect is that Rebel moves much to fast, factor 2-3. > >There are 2 cures: > >1) When the game starts use the "game in xx:yy" level only. You found out >yourself, well done. Yes, and I was glad that Rebel took much time for each move which I didn't really expect. Several times took more than 10min for a move. >2) When you (say) at move 40 want to change the time control to "game in >xx:yy" use the formula: (remaining time on the table clock * 2.5) - operator >time. > >Sorry for not informing you. > >Ed Jeroen told me about that but I didn't want to go any risk losing on time if Rebel really takes this time, e.g. g/130 while there are only 65min left. So in the first 3 games, well, I had to suffer more than I expected. BTW did you receive my e-mail with the games and Rebel's logfiles? Any idea what went wrong in which game? E.g. I am not sure whether the pawn ending in Rebel-Shredder is really won. Up to now I haven't found a way for a white win. Finally, thanks a lot for supporting me for this tourney. I hope you can see something from the games you can use to improve your engine.
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.