Author: Robert Hyatt
Date: 13:28:38 06/23/01
Go up one level in this thread
On June 23, 2001 at 10:27:18, Georg v. Zimmermann wrote: >I wrote about 2 pages then deleted them again to try to focus argumentation on 1 >point :) > >You ask why allow GMs to select the time controls against certain players but >not computers. >The answer is: because it is not the computer who selects the time control, it >is the operator! A human will most times play the time control he thinks he is >best at ( and if someone prefers 1 0 and the other 3 0 and they want to play >each other maybe they play 2 0, I have seen it happen often enough ). And your point would be? scrappy naturally plays any blitz up to 5 3. If a human wants more time, he has two choices: (1) play someone else; (2) ask me if it is possible to play something slower. For GM players I _always_ respond "sure". But most GM players prefer 5 0 and 5 3. Remember that Roman is playing 5 7 at _my_ request. He doesn't normally play anything that slow. > >A operator weakens the computers strength against humans considerable by not >setting formula to (inc == 0 && (time == 3 || time == 1 || time == 15)) ! >That in itself is ok like it is ok to have a special crafty account searching >only 5 ply. But it would not be ok to reduce that to 4 ply whenever it is >playing account "Tecumseh". Even though that would certainly make me happy ! > I don't see a thing wrong with doing that _if_ there is a reason besides rating manipulation at the core of it. In my case, it is a desire to see some "longer" games (roman has been playing 5 7 and 10 10 games quite a bit lately because I find they give more insight into how the program is doing that the fast tactical explosions do.) IE I have a _reason_ to play longer time controls against any GM if they ask. I want to see that whenever possible. I do _not_ want to see 2 12 and 5 12 games against computers or against weaker humans... It simply ties the program up for very long games (5 12 games can last an hour or more) and I don't get much useful information from the games. > > >Some more ideas about what you said below. Of course you have a point here. But >that doesn't mean that I don't :) > >The rating system would work much better if only tournament play would be rated >(like the FIDE system). ICC does that. It is called "the standard rating". > >But IMHO the rating system does not become invalid because I play some people a >bit more often than others. (One example: I am likely to play people who log on >in my timezone more often). > >The rating system does become invalid when I play a specific opponent a lot more >often than others. So maybe ICC should guard against that. I don't see how it becomes "more inaccurate". Ratings are not absolute, which seems to be the loop this discussion migrates to all too often. Ratings are simply a way to estimate probabilistic outcomes for playing other players with ratings. They have a significant error associated with them. The only way to get an _absolute_ rating would be a double-round-robin against _all_ players on ICC. Or in FIDE. Not likely to happen. So we accept an increased level of error and play far fewer games... But "invalid"? Never... This is simply about statistical probabilities, not absolute fact...
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.