Author: Howard Exner
Date: 19:47:48 07/21/98
Go up one level in this thread
On July 21, 1998 at 21:34:40, Danniel Corbit wrote: >On July 21, 1998 at 14:05:38, Howard Exner wrote: > >>Many posts on how to rank computer strength(or Anand's) contain >>some faulty reasoning. >> >> Dr. Tuerke's humour in exposing the flaw in Sean's Anand post >>was to me self evident. Yet others often employ similar reasoning >>when assessing computer strength. One goes like this: computers can't >>be all that strong because they get beat by IM X, Y and Z. Therefore >>X, Y and Z are stronger players than the computer. Now if X, Y and Z >>are also beating "other players" with more regularity than the computers >>are beating these same "other players", then yes X,Y and Z are stronger. >>And this will show up in their ratings over the large data pool of >>competitors. Is that the case? >> >>Otherwise, stealing a page from Dr. Tuerke, I will pronounce >>"Fischer was not all that great a player since he got clobbered lifetime by >>Geller". >I think the Anand/Rebel match is a pretty clear indication that at blitz chess, >computers are very, very strong. I think many people will believe I am just >being stubborn, but I do not believe that this shows Rebel is stronger than >Anand at blitz chess. Yes your example is correct in my view. Small data pools of results followed by lofty conclusions just doesn't seem scientific. These can be countered ad naseum with examples that show the opposite. We can conclude the obvious, that Rebel beat Anand in a fair 4 game blitz match and that Rebel beat Anand in a two game match of game/15. Now if Rebel were to continue cleaning up on a big cross section of strong players then generalizations would start making more sense. >Remember, the Rebel team had lots of opportunity to study >Anand, and Anand could not play against the new system in preparation. If he >played against the machine 100 times, he might discover a crucial flaw. This is reminiscent of GK vs DB. People said that Gary had this disadvantage also. I think Anand is very computer savy besides being the best GM in the last 18 months. I'm not sure what else he could have done to prepare. The computer's weaknesses are common knowledge now. Exploiting them may not be that easy as it used to at that fast time control. Yes the poor dumb beasts can be tricked into playing a poor opening system over and over and lose that way but what are the odds of that happening in such a short match? >That >having been said, it is also possible that Rebel 10, on that machine _really_ is >stronger than Anand. Lots of games will tell. The excellent showing of >computers on blitz games on the chess servers is good evidence that computers >really are very strong in this area. > >I must admit, I was very shocked by the result. My prediction was Rebel 10 >would get 1.5 points for the whole match, but I thought they were all going to >be longer time controls, and... and... the sun was in my eyes and the wind was >blowing the wrong way. > >Anyway, I think that whatever the real strength of the program, it would kick >the rear of 99.99999% of anyone going against it at short time controls anyway. >The fact that it did give Anand real serious trouble shows that it is quite a >program. > >King sized congratulations to the Rebel team. This was a very impressive >showing. This morning when I logged on to Ed's site and viewed the pgn printout of the first six games I kept double checking to see if I was getting the names and the result right. It was hard to believe. I'm curious what the coverage will be like in the larger chess community. > >What is the status of the longer time control matches?
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.