Author: Ricardo Gibert
Date: 01:33:30 07/31/00
Go up one level in this thread
On July 30, 2000 at 13:39:40, Peter Kappler wrote: >On July 30, 2000 at 02:03:07, Ricardo Gibert wrote: > >>What do you think would be the result of the following thought experiment: >> >>You get 5 games played by a player X rated 1500 against 5 different opponents. >>Their ratings range from 1300 to 1700. Now hand these games to a GM, but don't >>tell him the rating of player X nor tell the ratings of his opponent. Now ask >>the GM what he thinks is the playing strength of player X. >> >>I don't know about you, but my estimate of the rating could wind up being almost >>anything. I would have no point of reference for judging their playing strength. >>I would encounter long series of blunders by both sides in these games and would >>not have a clue about how to assign a number representing how strong player X >>is. I can't tell the difference between a 1400 and a 1600 rated player with any >>kind of reliability. They both get bad ideas, but one may get "better" bad ideas >>than the other. How do you rate that? >> >>The problem is perhaps I'm too high rated (2200+). Maybe if I was another >>C-player I could make a better estimate! Judging a peer may be more accurate? >>But then perhaps ego, 20-20 hindsight, etc. might come into play and the >>C-player would still be way off. I think a GM would have a tougher time than I >>would. How does a C-player think? That's a toughy. >> >>Impressions are a poor yardstick for estimating anything. An GM could be on the >>money, but could also be way off. His estimate would certainly vary with his >>mood, how much rest he got, rust, etc. just like his playing strength varies for >>the same reasons. >> >>There is no substitute for an objective determination. > > >You make a good point, and I do agree that the GM would struggle in the scenario >you described. > >But, I'm not sure why I handicapped the GM. He should start with the same >information as ELO -- the rating of the opponent! > >So I apologize for presenting a moving target, but I do believe that if the GM >is given the rating of the opponent, he will calculate a better performance >rating over a small number of games. > >That's it, no more concessions - I will stand firm, now. :) > >--Peter With the GM using the rating information, he is effectively performing a rough ELO calculation in his head, then tweaking the result one way or the other through his impressions, but the ELO calculation is doing the lions share of the work! In any case, I still think that bias will affect the result. If I were black or female, I would really hate a system affected my 1 persons impressions. Even if you could demonstrate it is better, the publics "impression" of your system may be negative, because of the _fear_ of bias. That may be paranoia, but I think it is better to minimize it with a system that is completely objective. Personally, I don't think it would be paranoia. There would be _some_ abuses. Yes?
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.