Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: cm pilz/faber v cm KK settings

Author: Detlef Pordzik

Date: 16:33:03 11/23/98

Go up one level in this thread


On November 22, 1998 at 08:13:07, James T. Walker wrote:

>On November 22, 1998 at 08:01:38, Dan Kiski wrote:
>
>>
>>
>>Has anyone else tried the pilz/faber settings v the Komputer Korner settings.
>>
>>My own tests indicate that the Pilz/Faber settings are stronger on all time
>>settings. All testing done using cm6000 on P233 MMX hash set at 24.
>>
>>I do find the Komputer Korner settings to be better than the standard cm
>>settings.
>>
>>Dan Kiski.
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>.
>Hello Dan,
>Can you give me the pilz/faber setttings for CM5500 ??  Maybe you can tell me
>where to find them and or a history of their origin.  What is their "Proof" as
>to being superior to the original CM5500 ??  I'm wondering if they are truely
>superior to the original CM5500 then why wouldn't the Mindscape people adopt
>them ??  This is new to me so please excuse my ignorance.
>Jim Walker
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>.


The history of this seach for better setting is way back when in late 1994 and
early 1995. Originally created at the time, when CM 4.000 was the standard
release of MINDSCAPE.
There were 2 parties working on this parallel - on one side, Mr. Walter PILZ,
and the other side were the 2 old, but very reputated comrades H.J. Schuhmacher
and Huber BEDNORZ ( Aachen and Colgne, Germany ). These gents were, BTW, the
founders of the valuable BT ( BEDNORZ / Freddy Thönnisen ) - and later
BS = BEDNORZ / Schuhmacher test series.
They were 1st given to publik in the Austrian Chess Computer mag PC SCHACH
1/ 1995 as the Version b.

Last metiones settings came out a couple of weeks earlier than the Mr. PILZ
settings.
These, however, did a good job in CM 5.000, not so good in CM 5.500 anymore and
sure are NOT close to the original gameplay of the engine in it's existing form,
2.61.
Here's the reason why :
CM 6.000, from what I've tested, roots back noticeable to the fomer gameplay of
CM 4.000 - which means, it's designed to play very active against the opposite
King, weak pawns and weak fields.
The setting " King safety 170 " slams on the brake of this engine, it's far too
high, the engine is advised by setting, to play more " carefully " - and this
contradicts to the origin. So, it somehow plays " constructed, unnormal " moves,
so to say.

For your convenience, I repeat the original BEDNORZ settings :

Attacker / defender +3; Strenght of Play 100; Randomness of Play 0; Book depth 0
Selective search 10, Contempt for Draw 0,0
Positonal :
Materail / Position -3; Control of Center 103; Mobility 103; King safety 133,
Passed pawns 100, Pawn Weakness 133
Material :
Own Queen 9,2 (!)
Opp.Queen 9,0

rest original

This, to prevent exchange willingness, if it ain't a countable benefit.
CM got no probs to exchange Queens early.....

I had them in thest series - they work out fine - and sure are worthy a solid
testing, if one is interested in this.

To ask something, one doesn't understand - and doesn't have the backgroung info
- is GOOD, never ignorant. THIS is the purpose of a NG, and I really wonder, why
all our approved experts didn't answer you.

Never mind :
as far as Mindscape is concerned, they just stick to a somehow " neutral
setting " originally.
Same as the original settings ttable_size=20 - in the remaining relevant .ini's
for this in CM 6.000, the cm.ini and the cmdefalt.ini.
This prevents, that somebody with little knowledge " kills " the prog by setting
too high values, comparing to his own system ressources.

If this is of interest for you, here's the way :
BOTH .ini's must have the same (!) value, of course.
The general rule her is :
1 count more means allways a doubling of the table size.
Example :
20 = 1, 21 = 2, 22 = 4, 23 = 8, 24 = 16, 25 = 32, 26 = 64 ( MB ).

CM is NO prog, who basically needs alot of hash tables, it is NO brute force
searcher.
For Blitz 4 MB. for tournament gameplay 40 / 120 I suggest 16-32 MB as just
fine, for overnight analysis 32 - 64 MB....allways keeping your RAM in total in
mind, to prevent swapping.

Hope, this helps

ELVIS





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.