Author: Uri Blass
Date: 01:22:16 08/09/03
Go up one level in this thread
On August 09, 2003 at 02:06:55, Jay Urbanski wrote: >On August 09, 2003 at 01:53:01, Uri Blass wrote: > >>>> >>>I think the best chance is to introduce a "Blunder Factor". The biggest >>>advantage the computers have (IMHO) is that they never blunder. Humans always >>>run the risk that they will miss something, and lose the game as a result. >>>Computers (at least within their horizon) don't have this problem. This is why >>>Kasparov said he accepted a draw in the last game against Deep Junior - he was >>>afraid he'd blunder again and lose the match. >> >>I do not believe kasparov. >>I believe that he wanted a draw because he thought that a draw is better than a >>win for him. > > >How would a draw be better than a win for him? because he gets another match and more money. The facts that I see after the match only increase that impression: Chessbase invites kasparov for another match against Fritz and they did not invite smirin who won the chessbase machines for another match. > >>>Against a human opponent if you blunder, there's always the chance that your >>>opponent will return the favor - especially if there is a chance for >>>counterplay. The computer will just eat you alive. So... I propose a Blunder >>>Factor.. a small chance of error on every move. It would certainly make for >>>more inreresting games against humans. I'd love to see some of these monsters >>>fight back from a stupid mistake. >>> >>>Don't ask me how to implement this.. just thinking out loud... >> >>In that case I see no competition. > >I don't see why a slight chance for error = no competition. because from the programmer point of view competition means that he does his best. Forcing him slower hardware mean competition but forcing him to play mistakes means no competition. > >>The question as I understood it was how to continue the competition between >>humans and machines. >> >>If the programmers do not compete then I do not see the point of it unless the >>target is to have better estimate of rating for humans so we can have computers >>with fixed rating. > >I didn't say don't compete. I just maintain that battling a machine that is >known to have a possibility of error would be more attractive to most humans >than one that never falters. I don't propose crippling programs - the chance of >a grandmaster blundering is pretty small, but it exists. And I certainly don't >think a blunder factor need be included in computer-computer matches. Telling the program not to do its best against humans is crippling the program. There are a lot of chessmaster personalities who do errors on purpose but if they play in tournament then Johan de koning(programmer of chessmaster does not do his best against humans so I can say that he is out of the competition against humans. > >>Today if all humans improve by 200 elo the rating is not going to change. >>Adding machines with fixed rating can change it. >> >>Uri > >Not sure I follow you. My point is the only reason to have machines who do mistakes on purpose is to have better rating system to compare between players of today and players of the future. Today we cannot do it by rating because if all players improve and continue to do the same results in games between humams then the rating of humans is unchanged and if there is a machine with fixed rating then we can see if suddenly the machine performs significantly worse and increase the rating of humans because the rating of the machine is unchanged.
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.