Author: Uri Blass
Date: 03:09:53 06/14/02
Go up one level in this thread
On June 14, 2002 at 04:24:23, Bas Hamstra wrote: >On June 13, 2002 at 19:57:43, Uri Blass wrote: > >>On June 13, 2002 at 17:49:26, Bas Hamstra wrote: >> >>>On June 13, 2002 at 17:12:06, Peter Berger wrote: >>> >>>>On June 13, 2002 at 16:01:37, Bas Hamstra wrote: >>>> >>>> >>>>>Of course there are a lot. Differences between stronger engines are very small, >>>>>see Leiden 2002. TheKing had slightly outdated hardware, and bang, it sinks to >>>>>the middle of the field. And numbers 1 and 2 in Leiden ran at significantly >>>>>faster hardware than the rest, coincidence? Give a strong amateur 2x time >>>>>advantage and the pro's will lose the last bit of advantage they have. Not many >>>>>amateurs can beat TigerPalm? Are you kidding? How about 30? Heck, in Leiden >>>>>Comet rolled 3 pro's up in a ball while NOT being on faster hardware: GT and >>>>>TheKing and Shredder :-) >>>>> >>>>>Best regards, >>>>>Bas. >>>> >>>>Uri is usually very careful with his wording ;). He said that there are few >>>>programs that can play on a 486 computer and beat Chess Tiger on the Palm. >>>> >>>>Afaik there aren't 30 amateurs who could do that, or do you disagree? >>>> >>>>Kind regards, >>>>Peter >>> >>>Well, let's say 20, and then I can list em for you :-) >>> >>>Yace, Crafty, Ferret, Callisto, Nightmare, SpiderChess, Monsoon, Bringer, >>>Nejmet, Tao, Insomniac, Comet, Patzer, Xinix, IsiChess, Anmon, GLC, Quark, >>>Queen, Amy, Amyan, Pharaon, Pepito. I am sure I forgot a few. >>> >>>Best regards, >>>Bas. > > >>I think that you overestimate the amateurs. >>I do not expect Nejmet on 486 to beat palm tiger. >>486 is about 3-5 times faster than the palm if I remember correctly but I am >not sure if I remember correctly the speed difference. >>I guess that tiger is about 5 times better than Nejmet and it means that if you >>give Nejmet hardware that is 5 times faster than Tiger the results may be >equal. > >We have had similar discussions about strength/hardware. I say top programs are >nowhere against a 4x faster stronger amateur. You say this is only worth 100-140 >ssdf ELO, not enough to compensate. I don't believe the 50-70 is a law of nature >that you can use to predict outcomes at totally different time controls. At >blitz it is probably more than 70, ditto for slow hardware. And even if we >agreed on the 70 extra ELO for 2x speed it remains to be seen if THUS 4x speed >means 140 ELO. In other words, I think your calculations are as good as >worthless. > >The pro's are *very* good, almost good enough to compensate for a factor 2. But >believe me, there it stops. > >Best regards, >Bas. If you are right then this means that the pro's are not very good and have not a big advantage relative to the amatuers. I do not know about evidence that crafty is better than the proffesionals if you give it 2 times faster hardware(I remember rating list when the difference between crafty and Fritz7 in blitz was clearly more than 100 elo on equal hardware so even if you assume that doubling the hardware gives 100 elo at blitz it is not enough for Crafty to be at the same level as the other amateurs. I also believe that most of the amaturs are clearly weaker than Crafty. Movei is at the level of the average free amateur and I believe that it is more than 5 times better than tscp(clearly not the weakest amateur and there are clearly weaker programs like Nero that is also not the weakest). I believe that movei can beat tscp in most of the games if you give tscp hardware advantage of being 5 times faster(I have a parameter that make movei play x times when you can choose x). I say that I believe becase I did not do enough games to be sure but I also believe that the same movei can lose against other programs when you give movei hardware that is 5 times faster(at equal conditions of 5 minutes/40 moves on p800 movei lost against list 31-4 with 5 draws and it also lost against amyan at 2 minutes/game 35-5. I told Movei to try every possible legal move as first move with white and with black because Movei has not a real book and I am happy with these results because previous versions of Movei did worse. I think that there is a big gap between different amatuers and I do not see a reason to assume that all the top 20 amatuers are almost at the same level. What are the results that support this theory?(I would like to see some rating list when every program appears twice in 2 different hardwares to prove your theory) 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.