Computer Chess Club Archives


Search

Terms

Messages

Subject: Chessmaster Personality grand test...

Author: Chris Tatham

Date: 14:02:24 01/18/06


Hi All,  :-)

For my own entertainment I'm currently running a large Chessmaster tourney to
establish to my satisafaction whether in a consistent test 'leading' CM
personalities are indeed stronger than the default settings and if so which is
the best one.

Details are as follows:

Test:

For both the CM9000 and CM10th default settings play (initially) 1000 blitz
games to establish benchmark performance. Repeat the same test with the same
conditions and opponents for each of the selected 'best' personalities (one from
each of the CM expert testers).

Personalities have been selected with the help of Graham Banks (thanks Graham!)
and are as follows:

CM9000
Apex (Graham Banks)
Default
Gladiator (Sedat Canbaz)
Mapi (Markus Pillen)
Mv2.5 (Milorad Madjar)
R1 (Wilhelm Hudetz)
Slayer 2b (Jason Kent)
Surak 8c (Axel Schumacher)
TO5 (Kurt Utzinger)
Wizard (Wael Deeb)

CM10th
Berean 5.54 (Jason Kent)
Cell (Dom Leste)
D2 Alos (Dimitris Poulos)
Default
Milan 2.3 (Milorad Madjar)
Ogre (Graham Banks)
Schumacher (Axel Schumacher)
SoFar 2 (Dan Ellwein)
WoDra+ (Wolfgang Draeger/Roland Scharmer)
Xperience (Wilhelm Hudetz)

If any of the CM experts wishes to change nominated setting please let me know
(testing will start later this week).

I am also toying with the idea of entering a further personality 'Lucky Dip' - a
selection comprising elements from each of the above as chosen by my kids!



Opponents:
(40 games each against 25 different Chessbase/winboard/uci opponents -
importantly CM personalities will NOT play each other!).


Test settings:
As a fomer CEGT tester, I have opted to keep the same CEGT blitz
conditions/settings for my own tournament


Results to date:
Benchmark of 1000 games for CM9000 & CM10th defaults has been set (subject to
replay of a small number of duplicate games 9 each (ie < 1%)).

Interesting so far is that CM10th (under these conditions) seems no stronger
than CM9000 and possibly even a little weaker....


CM9000 Default  525.5 / 1000   (matches won (15) drawn (1) lost (9)
CM10th Default  496.5 / 1000   (matches won (13) drawn (0) lost (12)


Testing of personalities will start later this week.


For those interested, full details of results and testing conditions are
available on my website - http://uk.geocities.com/bloodhound66@btinternet.com/


Thanks again to Graham Banks for his help and support with this test!

Best wishes


Chris



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.