Author: Tim Foden
Date: 05:17:38 12/03/03
Go up one level in this thread
On December 03, 2003 at 07:27:10, Uri Blass wrote: >On December 03, 2003 at 06:34:05, Tim Foden wrote: > >>On November 30, 2003 at 17:29:49, Jorge Pichard wrote: >> >>>Back in the good old days when all the Affordable Home P.C. were monoprocessor, >>>the best Chess programs were combined with the best available single Processor >>>P.C. in order to have a chance of winning the World Computer Chess Championship. >>>But now we don't even give recognition to the Winner of the Monoprocessor >>>Program which was GREEN LIGHT by Tim Fodem. This was a great accomplishment, >>>since in order to win you have to at least win against >>>one of the program using much powerful system such as Dual Opteron. Both El >>>chinito and Green Light won a game against Deep Sjeng. In my Opinion List was >>>going to upset another of the big boys besides BRUTUS, but unfortunately it was >>>not given the chance. >>> >>>PS: Congratulations to Tim Foden for having the courage to enter knowing that >>>you were going to be facing Top Chess Programs using Super Computers. >> >>Thanks. >> >>Well, I can't say it took any courage I'm afraid. :) I was expecting to be >>fighting not to be 16th, rather than to be placed 5th, so it is kind of the >>other way around. :) >> >>Cheers, Tim. > >Yes >It is a good surprising result espacially considering the fact that >greenlightchess3 is weaker than Crafty based on Leo's tournament result. I agree. >I do not believe that small changes in the evaluation can be a significant >improvement, True, but I didn't just change the evaluation. I haven't started a test gauntlet yet, but I ran a match overnight on my main machine, and 3.0.3.4 beat 3.00... pgn release date: Aug 23 2003 12:42:09 # Name Points Games Score +ELO Summary ---------------------------------------------------------------------------- 1 Green-Light-Graz-03r11 26.5 46 57.6% +53.3 (+19=15-12) 2 Green-Light-3.00-Default 19.5 46 42.4% -53.3 (+12=15-19) It may also be due to the opening book which has changed too. > so it seems that except luck the opponents were not as strong as >could be expected Yes. In such short tournaments luck is always a factor. GLC was lucky to draw against List. It was also lucky to draw against Jonny (which had a null-move zugzwang problem in one significant position). It was lucky to win against Ruy Lopez. >and there were 4 participants that were one level above the >rest(Shredder,Fritz,Junior and Brutus) > >I do not know what happpened to Deep Sjeng in the tournament. At one time Gian-Carlo thought that there may have been a bug introduced, something to do with the 64 bits. He was using a 64Bit version of WinXP with (I assume) Microsoft's 64 Bit opteron compiler. He decided he didn't trust this late in the tournament (he did node count comparisons and got different results between the 64bit and 32bit versions), and went back to 32 bits. After this I think Deep Sjeng was simply unlucky. Cheers, Tim. >The programmer claimed that there are improvements relative to Deep Sjeng1.5 >and I expected the probability of it to finish behind Greenight as very small. > >Maybe GCP can explain what happened. > >Is the new version of Deep Sjeng better than Deep Sjeng1.5 and Deep Sjeng had >extremely bad luck or maybe the changes made it worse? > >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.