Author: Martin Müller
Date: 01:36:47 07/10/01
Go up one level in this thread
On July 09, 2001 at 16:26:04, Peter McKenzie wrote: >On July 09, 2001 at 15:32:56, Robert Hyatt wrote: > >>On July 09, 2001 at 13:11:20, Ricardo Gibert wrote: >> >>> >>>I think there are many good reasons for not rating comp games on ICC: >> >>I have a slightly different view... I think it would be better if the only >>comps allowed were the ones run by the program authors. That would cut out a >>lot of "computer glut" on all the servers, and really minimize this problem. >> >>One computer (Crafty, say) won't influence the rating pool nearly so much as >>50 or 100 will, even though I do make changes frequently. >> >> >> >>> >>>(1) The comp accounts adversely affect the rating system on ICC, since: >>> >>> (i) The operators of the comp accounts change hardware (usually faster) that >>>makes the comp rating a misrepresentation of the accounts playing level. >>> >>> (ii) The operators of comp accounts will sometimes change engines, which >>>makes the accounts rating misrepresentative too. >>> >>> (iii) The comp operator may be the programmer, who may make changes to the >>>program, which again makes the accounts rating misrepresentative. >>> >>>(2) It becomes problematical for human players to gauge their progress with >>>respect to several months earlier due to (1), since ratings from different time >>>periods become more difficult to compare due to the adverse impact comps >>>probably have on the rating system. This is true generally anyway, but I would >>>think the inclusion of rated comp games makes things worse. >>> >>>----------------------------- >>> >>>As things stand now, it would make sense if operators only played rated games >>>against humans and unrated games against comps. Then they would get to play >>>human players more often. The reason I say this, is because human players stand >>>to lose a lot of rating points against comps, since comp ratings are generally >>>deflated. I think most of the prospective human opponents avoid comp accounts, >>>because the cost in rating is too high. Also, a comp rated by games against >>>humans only, will get a higher rating and therefore attract higher quality >>>opponents (both human and non-human). >>> >>>Take a look at the 2 accounts Scrappy and Crafty. Scrappy is rated over 400 >>>points higher than Crafty, since it only plays humans. The program is the same >>>and the hardware is the same. It should be obvious that any human that plays >>>Crafty is an idiot when he could play the much higher rated Scrappy instead. The >>>human stands to lose a lot of rating points against Crafty if they play an >>>extended series of games. Against Scrappy, he does not have to worry about this >>>so much. >>> >>>All the comp accounts that play both humans and comps are essentially the same >>>proposition to the human as crafty is. Humans stand to lose a lot of rating >>>points against them too so there should be a tendency to avoid them in >>>preference to an account like Scrappy or another human. >>> >>>BTW, it is a mistake to think ratings don't matter on ICC, since your rating >>>affects the quality of the opposition you face on ICC. Playing against the best >>>quality opposition as possible ought to matter to anyone serious about >>>improving. >> >>Ratings are _bloody_ on ICC. :) Nothing is more important to some. >> >> >>> >>>------------------------------ >>> >>>There is a certain Curtis Williams that seems to have irritated a lot of people >>>with his policy of playing only humans with his comp account Spitfire. This post >>>may seem to support what he has done, but this is not my intent. Actually, what >>>I think people really find irritating is his motivation for doing so. He wants >>>to portray his account as the "King of the Hill" among comp accounts, when what >>>he has done has only created the appearance this is so when it is not. >>> >>>It's funny that he managed to delude himself with this fantasy, despite >>>employing an "artificial" means to do it. >> >>Again, I don't see the point of playing someone else's program on a server. It >>takes time, computer resources, and for what? I would no more run another >>program than I would race someone else's automobile. The construction is part >>of the fun. >> >>I've never understood all the comp operators on the servers. I doubt I ever >>will... > >Its not that strange. After all, you already do it! Well partly anyway, since >you run Eugene's tablebase access code :-) > >Think of a player on ICC as a chess playing system. There are many things you >can tweak: > >1) hardware >2) hash settings >3) opening books >4) evaluation/style parameters >5) server formula's and seeks >6) the actual engine (via the source code) > >For people who run chess programs they didn't write, it is still possible to >tweak the first 5 things. I'm sure its alot of fun too. In some ways its not >so different to what you and I do with our own chess engines. > >After all, these people have input to the chess playing system, and so can take >some responsibility and excitement from watching the results unfold. Sure they >didn't 'write the whole thing', but then you and I didn't write the >compiler/librarys/operating system either - but that still forms part of *OUR* >chess playing system. > >Sure, we developed a bit more of our chess playing systems than the >'non-authors', but is it really so black and white? Are we so different? > >Do you really struggle to understand them? > >cheers, >Peter Hi Peter, exactly my sentiments! Regards Martin another operator of non-selfprogrammed chessprograms
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.