Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: an idea for an experiment about checking the meaning of rating in ICC

Author: Uri Blass

Date: 15:01:29 07/24/03

Go up one level in this thread


On July 24, 2003 at 13:54:53, Russell Reagan wrote:

>On July 24, 2003 at 09:25:22, Uri Blass wrote:
>
>>The point is that the difference in rating between players should tell me
>>information about the difference in playing strength between them.
>>
>>If X gets 2400 at standard time control and Y gets 2100 at standard time control
>>but X gets 3000 at blitz when Y gets 2300 at blitz then something is wrong in
>>the rating if X and Y play the same at all time controls(by using fixed time per
>>move and not pondering I make sure that the strength of X and Y is not dependent
>>on the time control).
>
>I don't think this is a valid conclusion. It is possible that engine Y was
>created by its author to only play blitz. Or maybe it has a simple eval like
>piece square tables, and so it can get to 2100 at standard time control and only
>2300 at blitz. Maybe engine X has a lot more knowledge, so it plays
>significantly better at blitz.

You miss the point.

X and Y are based on the same engine of the same author.
They have exactly the same evaluation and exactly the same search algorithm and
the only difference is that X is using more time.

X is using always 3 seconds per move when Y is using always 0.1 seconds per
move.

X plays better than Y because they have no bugs that cause mistakes only at 3
seconds per move and there is no design decision to do them weaker if they use
more time.

>To show that your conclusion cannot always be true, imagine an engine that tries
>to play the worst move. At 0.1 second per move, it will choose _better_ moves
>than at 60 seconds per move, because as it uses more time, it will be able to
>find worse moves.

I assume that the engine is designed to play better moves with more time.

Almost every engine (Crafty,Comet,Movei,Gerbil,tscp) is right for my purpose
except the possible problems of losing the same game again and again.

My point is that the difference in rating between Crafty(3 seconds per move)and
Crafty(0.1 second per move) is probably going to be bigger in blitz than in
standard time control and it does not make sense.

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.