Author: martin fierz
Date: 21:16:21 10/25/02
Go up one level in this thread
On October 25, 2002 at 08:43:12, Andreas Guettinger wrote: >On October 25, 2002 at 08:25:48, Uri Blass wrote: > >>On October 25, 2002 at 08:04:37, Andreas Guettinger wrote: >> >>>On October 25, 2002 at 07:49:17, Uri Blass wrote: >>> >>>>On October 25, 2002 at 07:43:52, Andreas Guettinger wrote: >>>> >>>>>On October 25, 2002 at 07:12:38, Uri Blass wrote: >>>>> >>>>>>On October 25, 2002 at 06:51:45, Sune Fischer wrote: >>>>>> >>>>>>>On October 25, 2002 at 06:29:07, Uri Blass wrote: >>>>>>>>No >>>>>>>>I did not think only about crafty. >>>>>>>>There were other cases when programmers released versions with no improvement or >>>>>>>>at least it is not clear if there was an improvement: >>>>>>>>Gandalf,Nimzo,Mchess,Genius >>>>>>>> >>>>>>>>Uri >>>>>>> >>>>>>>Based on eng-eng matches I guess. >>>>>>> >>>>>>>-S. >>>>>> >>>>>>I think that most people care only about them. >>>>>>I do not care about games against humans because it is clearly only a question >>>>>>of time until computers beat humans at all time controls. >>>>>> >>>>>>comp-comp is the interesting struggle and being better in comp-comp can also >>>>>>help to get better results against humans. >>>>>> >>>>>>Computer also can play for the same ideas that humans play so being better at >>>>>>beating them means in most cases also being better against humans. >>>>>> >>>>>>There are programs that can play for king attack and I do not think that you >>>>>>need humans to see the problems of your program against king attack if you have >>>>>>these problems. >>>>>> >>>>>>You can let your program play against sjeng. >>>>>> >>>>>>Uri >>>>> >>>>>I disagree. Better performance in eng-eng matches is no guarantee that a >>>>>programm performs better in matches against humans. >>>>> >>>>>And it leads the whole computer chess development into the false direction, >>>>>with no new concepts. >>>> >>>>I see no reason why not. >>>>New concepts can be productive to beat chess programs. >>>> >>>>Uri >>> >>>Look at the Kramnik-Fritz match games 2 and 3. In tactical positions, with lots >>>of pieces and queens, chess engines are not bad today (and this is indeed an >>>achievement of eng-eng matches). But in the other positions, the engines have >>>not a clue, are chanceless, and this will not improve from eng-eng matches, >>>because all the engines have this flaw. >> >>I do not think that all of them. >>There are engines that do not play a3 in game 3. >> >>Other engines have different holes in their knowledge so I do not think that >>Fritz has less knowledge than them. >> >>Uri > >I do. Fritz = fast, not much knowledge. Hiarcs8 = slooow, a lot of knowledge. that is a bit a simplistic statement :-) i don't mean to say you're wrong, but for example, fritzy is written more or less completely in assembler by people who know the ins and outs of modern processors. this is certainly not true for all top programs, but it is unfair to say fritz=fast <=> fritz has no knowledge, because it's speed is partly explained by great programming. >Unfortunately, I don't own Fritz, so I can not prove that. But I think it's >obvious, and nothing anybody will say will change my mind. :) just a thought: fritz played h4 in game 1 against kramnik, allowing a blocked position. hiarcs had outplayed smirin as black, but then allowed a fortress. mmh, i don't see any more knowledge there :-) finally, you can write a program with LOTS of knowledge and it plays like crap because it's useless/wrong knowledge. aloha martin PS: of course i agree with your post in general though :-)
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.