Author: Uri Blass
Date: 02:30:17 03/15/02
Go up one level in this thread
On March 14, 2002 at 23:58:13, Joshua Lee wrote: > >>Yes a test can be devised;now however I would recommend BS 2830.You should also >>note that the top pogrames are very close in Strenght.And finally if you could >>accept computer as playing at 2700 elo(which I think you dont seeing your posts >>here ) it would help you own accessment immensely >Well i can accept that they are playing over 2400 strength as far as a Tactical >Type position, Positionally that is really hard to say just look at the losses >that Deep Junior on a freaking 8 way system, and look at the Games GM Van Wely >won against Rebel Century. At best in those 4 games just mentioned a computer >played 400elo less than the human and HOW badly the computer looses when they >do. I would except if i were you that Computers Do not know everything a GM >knows , when they can Play Positionally , when they don't even need an opening >book, when they don't make the wrong moves immediately when they get out of >book, when they leave very few test set positions unsolved. THEN you can think >about how strong they are. You probably think the SSDF list is totally accurate. > >Yea Fritz is as good as Capablanca.... NOT. > >Tactically yes Computers can play as well as some GM's , Positionally , and >Endgame and even in the openings(if the comp is without it's book) no way. > >If a comp just wins and draws against an average opponent with an elo of around >2430 , doen't make the Comp a GM. > >Your basis on Programs being stronger is on many short matches with different >programs on different hardware instead of Lenthy matches which make the margin >of error much smaller. > >As an example GM Wely can lose to a program that is within 400elo of himself >and That same program can beat GM Wely. > >I must be a master because i've beaten masters at Blitz (2335elo). No, but some >of the games were 30 to 5 minutes 5 to 2 minutes , 5min, etc >I am not quite expert strength, I have Drawn WeakComp on ICC so I must be a >Master Like Every Computer is a GM. > >Your logic is based on small matches. Get a Master to play 100 Games lets see >the score, probably the Computer would win. Get an IM to play and the score >would be alot closer. I think that a match of 100 games is the wrong test. A better test is to let computers to play different players. In tournament I usually do not play against the same player again and again. Computer proved that they can get performance of more than 2700 against humans and not only in one tournament. Deep Junior did it and later Tiger could do even better performance with one processor. How many humans can say that they got performance of more than 2700 in a tournament of at least 9 games. I believe that this task is harder than getting a rating of 2500 and if you consider the samll number of tournaments that computers played it is even harder than getting 2600. I expect the best programs also to beat most 2600 players in a match 40/2 hours if the 2600 player cannot control the machine(it means that the programmer has the right to change the personality of the program between the games) I think that it may be more interesting to see matches with longer time control(2.5 hours/40 moves was played in the past and I see no reason not to use that time control in games against machines) We want to give the humans more chances and the general opinion says that humans can perform better against computers with more time (I remember that I read that 2 hours/24 moves was used many years ago) If we want to have some limit for the time of the game in order to finish the game in one day we can use 6 hours per game that is clearly slower than the time control that is used today. 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.