Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: I remember that DB was allowed to change params during the match...

Author: Mike Byrne

Date: 18:36:10 10/09/02

Go up one level in this thread


On October 09, 2002 at 16:26:26, Robert Hyatt wrote:

>On October 09, 2002 at 14:52:43, Terry McCracken wrote:
>
>>On October 09, 2002 at 12:50:22, Robert Hyatt wrote:
>>
>>>On October 09, 2002 at 12:46:42, Otello Gnaramori wrote:
>>>
>>>>...the question is if Fritz team is allowed to make the same changing of the
>>>>params of the program to tune it better against the Human Champion strategy, or
>>>>at least to make Fritz to not repeat the same moves each time...
>>>>
>>>>w.b.r.
>>>>Otello
>>>
>>>
>>>I think the pre-match agreement prevents this, which is a pretty stupid
>>>stipulation IMHO.
>>>The human can study a game and change his strategy for the next, but the
>>>computer cannot?
>>>
>>>The agreement is really a joke, IMHO.
>>
>>Do you think it would really make that much difference, considering how Kramnik
>>lures it out of book early and then proceeds with his strategy to exchange
>>Queens, to exploit endgame weaknesses all programmes suffer from?
>>
>>If you change parameters for the programme to hold it's Queen, for example,
>>it may be a serious mistake, allowing Kramnik a strong or even winning
>>advantage.
>>
>>I think we can't compare changing parameters with Fritz, a commercial programme,
>>on an 8 CPU machine to that of Deep Blue a supercomputer, which has tremendous
>>flexibility. I just don't believe there is room for serious improvement in this
>>case.
>>
>>IOW it may make little difference or worse, causing Fritz to go down in flames!
>>IMO Fritz is playing well and fighting well, considering it is drawn into the
>>area of the game all programmes struggle with.
>>
>>Terry
>
>
>That I can't answer.  But the rule saying it can't be done is just bad, period.
>IE I have
>some global sort of evaluation terms that I could (and would) modify between
>rounds.  If
>I noticed the human trying to lock up the pawns, I'd turn that scoring term up a
>bit to try to
>avoid it.  If he tries to attack then I'd turn that up to push him another
>way...
>
>Fritz is playing fine, for a computer.  Unfortunately, it _needs_ to be playing
>like a GM,
>which it is not doing at the moment...
>
>Kramnik's match agreement highlights this problem with all the practice he has
>had against
>an opponent that he _knows_ can't change ideas between rounds...

I have to agree with you - it's a bad rule.  I am a 1600-1700 player -- I can
now regularly (after 2 years) beat Chess Genius Palm (often peg @ 2000) ona
regular basis - it doesn't change - I know it's weaknesses.  I am not a 2000
player now ,  I'm just a 1600/1700 player that can beat Chess Genius Palm.
Under these conditions - access to the program for a long before the match and
with no changes once the game started - the program would have to be several
hundred point strongers than the GM to have a chance.

But I don't think the Fritz people really cared - it's really not about winning
for them - like it was for IBM.  This is about computer chess promotion - and at
that, they succeeded 100%.  There is a lot more interest right now about
computer chess than there's been for some time, since the DB-GK match.  IBM's
plan all along (I believe) was to beat GK and then to take the box apart.  It's
like hitting a home run in your last at bat.  There will never be the same
excitement regarding human/computer chess as there was in 96/97.  Look at
computer checkers - although never having the interest of computer chess -  the
high point was when Chinook won it's first WC against a human.  It's been a PR
non-event after the first couple of wins aginst top humans.



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.