Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: CM8 Personalities at 60 in 10 (Legend,8999,Fun,Ming,Pillager,Ace)

Author: Steven J. Brann

Date: 05:10:00 05/27/01

Go up one level in this thread


On May 26, 2001 at 00:50:53, Graham Banks wrote:

>All 16mb hash.
>Each personality played 4 games against the others.
>
>Legend      - 14.5 / 24
>8999        - 13.5
>Fun         - 12.5
>Ming        - 11.5
>Pillager    - 11.0
>Bendorz     - 11.0
>Ace         - 10.0
>
>Each of the above personalities will be added into the tournament that I am
>running involving 15 other personalities.
>For those interested the settings of Legend, Ming, Pillager and Ace are below.
>Graham.
>
>Legend   - A/D=0 SS=12 M/P=0 CC=120 M=110 KS=160 PP=100 PW=130
>         - Q=10.0 R=5.4 B=3.4 N=3.3 P=1.0
>
>Ming     - A/D=0 SS=12 M/P=0 CC=120 M=100 KS=170 PP=100 PW=140
>         - Q=10.2 R=5.5 B=3.4 N=3.4 P=1.0
>
>Pillager - A/D=0 SS=12 M/P=-10 CC=110 M=105 KS=160 PP=100 PW=115
>         - Q=9.5 R=5.3 B=3.2 N=3.2 P=1.0
>
>Ace      - A/D=0 SS=12 M/P=0 CC=110 M=120 KS=150 PP=100 PW=130
>         - Material values same as for Legend.

Hi Graham

When I created CM8999, I was uncertain as to the SS setting that would work best
with varieties of hardware speeds and time choices.  I've found that the
original SS=7 works pretty well at short time controls as you have found here.
But at longer time controls performance wanes significantly.

Now, I'm trying to find the optimal SS setting.  Currently I have it set at
SS=11 with the following results:

All games PIII 933Mhz, 16MB hash

58 games at game/5:

CM8999       30.5  (Holds its own but sticks with the pack)
CMFun        30.0
CM8Utzinger  26.5

But contrast that with 42 games at Game/60

CM8999       25.0  (Much more promising - Would have been worse with SS=7)
CMFun        19.0
CM8Utzinger  19.0

I'll continue with many more games under different time controls to optimize,
(including a control with Defult CM settings - 16MB hash).  It appears that one
of the greatest influencers of results is the selective search setting.

This leads me to wonder the degree to which chess programmers vary the impact of
settings in real-time based on the game at hand, speed of hardware etc.  Any
thoughts?

Steve




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.