Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: But Not Yet As Good As Deep Blue '97

Author: Robert Hyatt

Date: 10:50:06 07/21/00

Go up one level in this thread


On July 20, 2000 at 14:43:10, Chris Carson wrote:

>On July 20, 2000 at 14:30:41, Robert Hyatt wrote:
>
>>On July 20, 2000 at 14:12:32, Ed Schröder wrote:
>>
>>>
>>>Never have we seen these 40 games. The least one might expect that data
>>>can be checked if it comes from a commercial source don't you think?
>>>
>>>For the rest of your arguments I think Chris did a very good job.
>>>
>>>Ed
>>>
>>>
>>
>>
>>Not seeing them doesn't bother me at all.  Had I known _nothing_ about
>>deep thought, and the later DB hardware, I might have a different opinion.
>>But I _know_ what deep thought did to _everybody_.  I have no trouble
>>believing that a new version of that processor would be even stronger...
>>
>>IE their past performance showed they were _clearly_ the best there was,
>>computer-wise.  Which makes taking their word about the 40 games pretty easy
>>for me to do...
>
>I am still waiting on what programs were used and what the rules and
>set up was and who the operator was.  I do not think that any of the
>programmers had approved this match (thus the operator).  Even if this
>is a true result, it is not valid with out full disclosure ahead of time.
>
>You critize people all the time for "program x beat crafty 100 to
>nothing" and demand to know what the book, settings, pgn's ... in
>this forum.  Why the double standard.


There is no double standard at all.  There are people.  And then there are
people.  IE if Ed played games using crafty, my first impression would be that
he would do it right.  Because he would recognize no pondering, no book, lots of
paging, etc.  If someone unknown to me plays games, I would ask questions
because most of the times there are flaws in the setup... sometimes the flaws
favor crafty, sometimes not.

Murray ran the games.  He is _very_ experienced in computer chess.  If he told
me (and he did) that they beat Crafty just as badly as they beat the other
programs, I wouldn't (and didn't) bat an eyelash, because there was no doubt
in my mind that he ran the test with a good configuration for Crafty and the
others.

I have known him for 20+ years, and over that time period I have learned to
trust him.  I _still_ do.

In light of that, does there still seem to be a 'double standard' at work
here???





>
>Did they test against Deep Junior?


There was no deep junior.  Do you feel that deep junior is significantly
stronger on the 8-way box than it was on the 4-way box it used in the last
ICC chess tournament?  If so, why?  They tested against the commercial programs
running at an estimated 200K nodes per second (one chess processor, crippled
for debugging/testing).  Do you not think that running 1,000 times faster would
more than offset any gain DJ would have on the 8-way box?  I do...  DJ probably
barely gains 2x from 4-way to 8-way.  Probably 3x from 1 cpu to 4.  They gained
a factor of 1000 if they used the _real_ DB hardware...

old news, of course...






>
>Best Regards,
>Chris Carson



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.