Author: Robert Hyatt
Date: 10:58:39 07/21/00
Go up one level in this thread
On July 20, 2000 at 14:55:52, Chris Carson wrote: >On July 20, 2000 at 14:52:23, Robert Hyatt wrote: > >>On July 20, 2000 at 13:01:58, blass uri wrote: >> >>>On July 20, 2000 at 10:58:34, Chris Carson wrote: >>> >>><snipped> >>>>I do not expect you to ever admit a mistake or that someelse >>>>might have a valid point. I have never seen that. I do see that >>>>96 DB is a dead issue and you try to prove 97 DB 90% supriority >>>>based on DT vs 386 and 486 machines. This is just plain not valid. >>> >>>I agree. >>>Hsu did not want people to believe that Deeper blue has 90% superiority. >>> >>>Hsu did not want people to believe that Deep thought can get more than 90% >>>against Fritz3. >>> >>>If we assume that Deep thought could get more than 90% against Fritz3 p90(at >>>tournament time control) then hsu could convince people about it by doing public >>>games between Deep thought and Fritz3. >>> >>>He could know that people have the impression that Deep thought is weak after >>>Deep thought lost against Fritz3 and drew against wchess(p90) >>>He did not try to prove to the public that they are wrong and previous games are >>>not proof because deep thought did not play against a lot of commercial programs >>>on good hardware and I cannot learn much from results against not commercial >>>programs that may lose in part of the cases because of bugs. >>> >>>The reason for the fact that Hsu did not try to prove that Deep thought was >>>strong may be one of the following: >>>1)Deep thought was weak. >>> >>>2)He wanted kasparov to believe that deep thought was weak. >>> >>>In the second case his behaviour was bad because he tried to use psychological >>>tricks to win kasparov(I do not think that kasparov could learn much about >>>Deeper blue from watching many games of Deep thought because Deeper bluer was >>>clearly different but kasparov could avoid wrong assumptions that lead to bad >>>preperation in this case). >>> >>>If he wanted kasparov to believe that Deeper blue is weak than he deserves that >>>people will think that deeper blue is weaker than it really was. >>> >>>I see no reason to believe the more than 90% against Fritz3(p90) when I saw no >>>proof for it. >>> >>>Uri >> >> >>I see no reason to doubt the results. Doing so is directly calling Hsu, >>Campbell, Hoane, etc "liars". I know them to be better than that... > >Not liars. Those are your words so own them! Not my word. I belive Hsu/Campbell. If you don't, you _must_ think they are liars. There is no in-between. They told the truth and you believe them. They made a statement and you don't believe it. If you are wrong, they told the truth. If you are right, they are liars. Pretty simple, really... > >We do not know what programs were used, settings, books ... Yes we do... the names and versions of the programs were specified by Murray when someone asked him at one of his lectures. I no longer remember, but then I don't consider it very important as it was not a surprise. I remember Genius was one of the programs. Perhaps the other was Rebel. Ed might remember, I don't. We know they used the engines as distributed on the commercial market, with the tournament books turned on, reasonable hash settings, everything else at default settings. Again, this is all old news. I didn't attend the lecture by Murry or by Hsu, but several that did reported on this. I didn't know they had played 40 games until way later. Hsu only originally told me about 10 games that were all wins by DB. Later we (someone asked at a lecture) found out that there were actually 40 games with DB single chip getting 38 points. > >When someone reports "crafty lost 50 to 0 against..." you insist on >knowing all the details... why the double standard? No double standard. If someone well-known does such a test, I assume it was done right. If it is someone I never heard of, I usually ask questions. That's all there is too it. I trust those I know well. quite natural in fact... > >Not a valid test, we do not know the programs, settings, operators, >no pgn's. You just use this becasue it fits your needs, not because >it was a valid test. It was simply an experiment they ran, in their lab, to see how their hardware evaluation changes were working. It wasn't intended to prove anything other than for their own testing purposes. I play lots of games against commercial programs. Have you seen me report the results anywhere? The only results I ever mention are the results everyone sees as Crafty plays on ICC in the public. My private tests are for just that... testing or debugging. Hsu and Campbell didn't consider it important at all, just interesting. When he casually mentioned it to me, and I relayed it to r.g.c, it created a lot more interest and surprise than either of us thought was justified. So it was done casually, but with care to make the test as difficult as possible for DB, because the purpose was to test the machine, not produce spectacular results. > >Best regards, >Chris Carson
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.