Author: Rolf Tueschen
Date: 06:28:22 02/16/03
Go up one level in this thread
On February 16, 2003 at 08:38:43, Uri Blass wrote: >On February 16, 2003 at 08:28:38, Frank Phillips wrote: > >>On February 16, 2003 at 08:19:08, Uri Blass wrote: >> >>>On February 16, 2003 at 07:23:29, Frank Phillips wrote: >>> >>>>On February 16, 2003 at 07:10:34, Uri Blass wrote: >>>> >>>> >>>>> >>>>>The gap between the best programs and part of the free programs programs is more >>>>>than 500 elo on equal hardware. >>>> >>>>Can you point to the data source that supports this claim. > 500 seems a bigger >>>>gap that indicated by SDF and ICC. >>> >>>There are a lot of rating list >>>The gap between the best free programs and another part of the free programs is >>>also more than 500 elo. >>> >>> >>>Look at the following list >>> >>>http://www.digichess.gr/infiniteloop/ratings/rapid_rating_il2r.txt >>> >>>Crafty17.9 2672 >>>Movei0.07a 2052 >>> >>> >>>Movei0.07a(my old program) is not extremely weak and it is >>>more closer to the top than to the bottom that is >>>LaMoSca 0.10 1183 >>> >>>I said that the gap between the best programs and part of te free programs is >>>more than 500 elo. >>>I did not talk about the best free programs. >> >>I did not understand your sentence. Clearly if the one you describe are bey >>definition those >500 worse then the statement is obviously true. I had in mind >>Riffian, Crafty, Ferret etc >> >>> >>>> >>>>> >>>>>Hardware that is 100 times faster will not be enough to compensate for that >>>>>difference. >>>>> >>>>>We do not know what was the level of the software of deeper blue because it >>>>>never played games on equal hardware. >>>> >>>>Why should it have to play on equal hardware? It was designed to a different >>>>paradigm. >>> >>>If people use the speed of it as a reason to convince people that it was better >>>than we need some information about it's level on equal hardware. >>> >>>> >>>>> >>>>>I also doubt the claim that the hardware was 100 times faster and we have no >>>>>proof for the number of nodes that it searched(200M nodes could be a >>>>>psychological war against kasparov when the real number was only 20M). >>>>> >>>>>Uri >>>> >>>>I wish IBM (or someone) would fund Hsu's next chip. Then we would finally be >>>>able to answer these questions Of course, some would then deny the facts if >>>>they did not fit their expectations ;-) >>> >>>No >>> >>>It can only prove in the best case for Hsu that Hsu is capable to do >>>a chip that is better than the top programs of today. >>> >>>I did not claim that he is unable to do it. >>> >>>It is not going to prove that DB97 is better than the top programs of today. >>> >>>Uri >> >> >>Moot as alawys. It was interesting listening to Fed. on chess.fm. He praised >>Junior, but did not have it in the same class as DB 1997... Of course he may not >>be impartial, but then again neither are the current set of commentators. >> >>My interest as stated before is more whether we actually need something as good >>as DB to stand toe to toe with the elite GMs, or whether a good software package >>(Hiarcs, Fritz, Junior....) on a fast PC is sufficient. >> >>If Hiarcs, Fritz and Junior really are as good as DB97 (in their different ways) >>then the answer is yes anyway.... :-) > >I believe in software and not in hardware. > >I believe that there are a lot of ways to improve chess programs significantly. >I believe that programs search too big trees. > >I believe that it is possible to get the same result by trees that are more than >100 times smaller even relative to the commercial of today if people get the >right ideas. > >Uri I'm still not _totally_ convinced that the idea of the Turk is an _absolutely_ bad idea. Rolf Tueschen
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.