Author: Rolf Tueschen
Date: 02:50:08 07/17/02
Go up one level in this thread
On July 17, 2002 at 01:38:45, Uri Blass wrote: >On July 16, 2002 at 18:50:26, Rolf Tueschen wrote: > >>On July 16, 2002 at 16:08:01, Amir Ban wrote: >> >>>On July 15, 2002 at 08:10:10, Omid David wrote: >>> >>>> >>>>I strongly believe that all chess programs are dump, not being able to see some >>>>of the obvious positional elements in a position. If a chess program like Junior >>>>or Fritz loses 5 games in 100 to me (2250 Elo), it means that computers are by >>>>no means superior to humans. (I certainly won't be able to beat a 2700 Elo Human >>>>5 times in 100 matches!) >>>> >>> >>>A 2700 Elo player is 450 rating points better and so should beat you about 10 to >>>1. >>> >>>If you score only 5% against a program, then its rating is higher than 2700. >>> >>>Amir >> >>First my gratulations for the title! > >I will also ad my congratulation for the title. >I always prefer to see a program from Israel takes the title. > >> >>Then to your statement. It's by no means reasonable. Perhaps the human >>chessplayer has difficulties with memorizing openings while you have the big >>databases form GM chess. >> >>In general you should always be aware of that chess strength is a matter of the >>games and not simple arithmetics. For me it's very troubling that you speak of >>more than 2700 Elo as if that could be possible for your program or any other. > >1)Amir did not say that the program's rating is above 2700 but only that if 2250 >player can score 5% against it then it is higher than 2700(the first assumption >does not seem to be correct). > >2)If you consider only games then old version of Deep Junior already had >2700 >performance against 9 GM's. > >I believe that GM's can do better(and the win of smirin against programs 5-3 at >faster than tournament time control suggest that they can do better) but the >2700 is not based only on fantasy. > >Uri You misinterpreted my phrasing. Let me explain right to the point. The mentioning alone of a Elo number for a commercial computerchess program of 2700 or even above 2500 is, in my eyes, an outrage against all what we know about chess. Let me add the usual reasons. ° books and tablebases DURING games (you shouldn't do this!) ° the 9 GM or any GM in a show act had _no_ sufficient preparation on anticomputer technology These two factors are the main reason for the actual confusion. Hence, because you do _not_ speak about chess (as we know it), you have no right to present Elo numbers. If I run with a sky rocket on my backside the 100 meters, the officials would not include me in the world record lists, since I cheated on the classical rules of athletics. It's so basic! Actually you and Dr. Hyatt too declare that successful play against computers is not the same as classical chess. So, by force, all results from show acts where famous GM played classical chess against doped machines do not justify the quick calculation of "Elo numbers" of such hybrids, to put it politely. Read Bob's sighs about his experiences on ICC where a few GM make short work (or putty) of "your" 2700 machines (of course all databases included)... ;) 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.