Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Rebel 8 - ELO level realism

Author: Don Dailey

Date: 13:25:19 09/17/98

Go up one level in this thread


On September 17, 1998 at 15:47:54, John Coffey wrote:

>I have strong opinions about this topic.   Let us assume that dropping the
>search by 1 ply equals 1 rating class, or cutting the number of nodes in
>half drops the strength by 80 to 100 points.  If that is the case then it
>should be possible to weaken most programs down to near 0 strength.  But the
>fact is that most programs have a lower limit that they will not go below
>unless you adjust other factors such as the value of the pieces or the "blunder
>range."
>
>If and when I write my chess program (starting next month), I intend to have
>a feature where one could set the rating at any level from 0 to whatever the
>upper limit is.   The limiting factor will be the number of positions that
>the program can look at, and other factors such as "blunder range" should not
>be necessary.
>
>John Coffey

But 0 is not the weakest possible strength.  You can play much
weaker than this.  In fact you can play much weaker than playing
a random legal move!

One standard way of weakening your program  is to truncate the
quies search of a 1 ply search.  This has the merit of looking
almost like a human blunder, where injecting a random move
occasionally just looks very silly and mechanical.


- Don



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.