Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Checkers: Las Vegas and Chinook

Author: martin fierz

Date: 20:40:12 09/09/02

Go up one level in this thread


On September 09, 2002 at 23:05:24, Russell Reagan wrote:

>So do you think that Schaeffer is acting in a similar way to IBM? Win a huge
>match, then "retire" it because they know the real truth about it's strength? I
>think that if Deep Blue would have played a dozen matches against the world's
>top competition, it would have lost more matches than it won, and then the aura
>surrounding Deep Blue wouldn't be what it is. Do you think the situation with
>Chinook is similar?
>
>Russell

hi russell,
hmm, good question and kind of awkward for me: i already got bashed for saying
what i did in my report, and if i were to say "yes" now, then i will probably
get bashed some more :-)

but well, to answer the question (to all of you who don't like what i say: this
is just what i THINK - it is only my OPINION - and it is *not* rude to have an
opinion!):

i think chinook was clearly stronger than any human player, after the deaths of
tinsley and lafferty, who were the world's strongest players. so no, in this
sense, the situation is not similar - chinook would not have lost any matches
against any competition at the time it retired.
my best guess from the very little i know about the program is that it is "just
another checkers program". many people here think deep blue was a really bad
program compared to today's micros, and i don't think that goes for chinook. it
would be interesting to compare it on equal hardware to today's programs. there
is nothing that makes me think that it is either much better or much worse than
the programs of today.
there are the two blunders in the match against lafferty, but 1. AFAIK in one of
the games, they had a problem with their computer and had to run on a weaker
machine, and 2. i think that this type of comparison is fundamentally flawed
(the "deep blue is crap" camp brings this form of argument up from time to
time): out of maybe a 200 positions where chinook could make a mistake against
lafferty in the match (where it was out of book and not yet in the database), we
select the two where it did make a mistake. chances are good that my program
doesn't make a mistake there (it doesnt). but chances are also very good that my
program would make a mistake in one of the remaining 198 positions where chinook
played the right move, so this type of comparison just doesnt work.

so there is absolutely nothing i can compare my program against chinook with -
unlike deep blue, where you have all the logfiles. comparing without logs,
without knowing how fast chinook was searching in a particular position (they
played on many different machines, depending on what was available, and
sometimes had to switch because of problems. also, they had different parts of
the endgame database at different times and i don't know what exactly they had
when.), what it saw and when it saw it, is totally useless. there is much less
to go on with chinook than with deep blue, and already there, opinions differ
wildly.

i guess that doesn't really answer your question :-)
 aloha
  martin



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.