Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: SSDF(Deep Fritz - Century 3)AMD K6-2 450, 3-1, now 18.5-6.5

Author: Ed Schröder

Date: 12:32:52 01/17/01

Go up one level in this thread


On January 17, 2001 at 14:42:15, Tony Hedlund wrote:

>On January 17, 2001 at 04:22:18, Ed Schröder wrote:
>
>>On January 17, 2001 at 02:09:13, Garry Evans wrote:
>>
>>>On January 16, 2001 at 12:35:20, Ed Schröder wrote:
>>>
>>>>On January 16, 2001 at 09:06:45, Jari Kylmälä wrote:
>>>>
>>>>>On January 16, 2001 at 06:19:40, Ed Schröder wrote:
>>>>>
>>>>>>New in Century 3 (being in auto232 mode) is that so now and then the
>>>>>>program checks itselfs by doing a 11 ply search (about 10-15 seconds)
>>>>>>from the start position. The total number of nodes is checked and if
>>>>>>they don't match (they should always match of course) the engine is
>>>>>>damaged and an error message is written in the log-file AUTO232.LOG
>>>>>>
>>>>>>Even if only ONE error would occur that is the ultimate proof the
>>>>>>autoplayer software damages Rebel playing strength as suspected since
>>>>>>years.
>>>>>>
>>>>>>The error message "engine damaged" can be found almost without exception
>>>>>>in the log-file after each autoplayer session.
>>>>>>
>>>>>
>>>>>How is this actually possible? Century or AUTO232 has a memory flaw and
>>>>>somehow AUTO232 is writing somewhere in the memory which should be
>>>>>allocated by Century? I don't have AUTO232 myself so I don't fully
>>>>>know how it's functioning or the protocol/interface it uses but this sounds
>>>>>very strange to me.
>>>>
>>>>Yes it is strange. The suspect is indeed the driver writes outside its
>>>>own memory. I have asked for the source code of the driver but it is not
>>>>available anymore.
>>>>
>>>>The other possibility Rebel hurts itself is very unlikely. Playing hundreds
>>>>of engine-engine games never returns the the error message.
>>>>
>>>>Ed
>>>>
>>>>
>>>>>--
>>>>>Jari
>>>
>>>
>>>
>>>
>>>
>>>  Actually I was very surprised at the SSDF result of rebel Century vs Fritz, My
>>>Buddy and I played a 20 game match, I have a 750 using Century 3 and My Friend
>>>has A Thunderbird 1000, the games were played manually , Century slaughtered
>>>Fritz 13-7 !!! and he had the much faster hardware!
>>
>>It has been reported to me by almost all people who still take the time to
>>play manual games that the results of Rebel differ significant playing
>>manual or auto232. It is just a confirmation of what I suspected first, now
>>proven by the introduction of the log-file. Also almost all manual events
>>Rebel participated (manually) till now have been successful with the exception
>>of the "Thueringen" tournament. But then in this tournament Rebel played with
>>an untested personality and also in some games the "Playing Style" was set
>>to "defensive" which is a disaster.
>>
>>Rebel Century 3 (manual) results (without Thueringen) are listed on:
>>
>>http://www.rebel.nl/r11-resu.htm
>>
>>
>>Most German testers got the message and if they include Rebel in a tournament
>>Rebel's games are played manually and I am most thankful for that.
>>
>>>Something is definitely
>>>wrong with ssdf testing.
>>
>>The SSDF people do a fine job only that they should start to believe producers
>>when they put something in the manual instead of calling them bad losers. But
>>apparently producers are all liars looking for poor excuses if their baby does
>>not perform well on their list. They don't even ask themselves the possibilty
>>if they (maybe) could be wrong in their judgement.
>>
>>Ed


Hi Tony,

>I have a similar auto232.log file that James Walker show us. And we do believe
>*you*

Thank you!


>(not what every producer put in there manual) that auto232 can hurt
>Century 3's performance. As Thoralf stated in his comments to the latest list.
>And if it hurt so bad as this match indicates, we might have taken a wrong
>decision. But did we/you know how much the autoplayer software damages Rebel
>playing strength?

The sad thing is that I really have no idea about the damage, only that
there is damage. I wish I could give you a more concrete answer as I am
of course also curious how Rebel would do in comp-comp as I still believe
in your concept playing hundreds of games.

I stopped hunting for the bug when I saw the results of the log-file and
have put my cards on the Rebel-Tiger GUI.

Keep up the good work.

Ed


>Tony



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.