Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: RebelRex (C) now on top ICC -- after 3-0 match against SWM(GM)

Author: Uri Blass

Date: 00:53:56 07/16/01

Go up one level in this thread


On July 16, 2001 at 02:19:22, Peter Berger wrote:

>On July 15, 2001 at 18:05:24, T. Dex wrote:
>
>>RebelRex is NOT setting any formula making it favorable for the computer.
>>RebelRex stands by its settings and program... RebelRex will play anytime
>>control the human wishes.
>
>I didn't mean to imply you ever changed your formula . Your formula is
>favourable for the rating of your account because of the way ICC ratings work .
>
>Typing "best C" and then "vars" of the ranked computer accounts looking at their
>formulas shows they all have set up something very similar.
>
>I understand the fun of watching computer programs play against strong titled
>players . It seems to get them to play you you currently need a rating of at
>least 3000 .
>
>So - let me try to describe the life cycle of a typical ICC computer account
>going for the top as it looks to me.
>
>a.) Start your new account . How do you get your rating up ? Have a look at
>other C accounts with open formulas , preferrably on slowish hardware and play
>them as long as they put you on your noplay list or you have accomplished your
>goal - best are some amateur programs . If you are more patient you can do b.)
>at once ; takes a little longer but should work nevertheless .
>
>b.) Once you reached 3000 - set your formula to something restrictive ( don't
>forget !computer ).
>
>c.) Enjoy watching your account going up on the ranking and watch some nice
>games .
>
>d.) Eventually you get bored and give up on it and go on with your normal life .
>
>I see nothing wrong in principle with this approach ( especially if you avoid
>the somehow immoral a.) ) - every one has the right to do that. If this does
>some harm it is not the fault of the operator but ICC's .
>
>It is somewhat annoying for the other computer accounts though that would like
>to see games against strong titled players, too . For most humans "program is
>program" - the ones who play them regularly in rated games think they can profit
>from their weaknesses and get rating points from them ( similar approach to
>a.)-d.) btw :) ) . In general the one who loses rating is the human though.

This problem can be solved by the following way:

Every program should have 2 rating numbers(unless it does not play both
computers and humans).

rating A list is going to be the rating that is calculated based only on the
games against humans(this rating list is going to include the humans).

rating B list should be the rating calculated based only on comp-comp games.

if the average of the rating of programs that are in both lists is not the same
then we should add a constant for the rating of programs in rating B list.

Uri



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.