Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: deep blue elo

Author: Robert Hyatt

Date: 08:36:01 10/19/99

Go up one level in this thread


On October 19, 1999 at 07:51:13, Graham Laight wrote:

>
>>>How about this:  "this output looks suspicious and really needs an explanation
>>>by IBM" after he had been told _exactly_ how normal this output looked.  I
>>>even posted _identical_ output from Crafty that showed exactly the same kind of
>>>'panic time usage'.
>>
>>The subject wasn't about panic time. The subject was about the DB main-line
>
>What is "panic time" in this context?


DB was looking at move X, and the score dropped iteration by iteration,
until the point where it became 'concerned' and entered what they call "panic
mode" which causes the program to take extra time to find a better move, since
the score has dropped so far.  Most programs do this.  But for some reason,
a few thought that DB's doing so was a cover-up for having a human intervene
and causing it to play a different move than what was 'best'.  It was baloney,
as the DB output clearly showed.  But the claim was made and then restated
multiple times.  The main problem was that it showed an eval drop, then went
into 'panic time' mode, then announced "root score restored" (which meant it had
found an alternative move that raised the score back to near where it was before
the previous best move dropped in score) and it promptly played this new move
without it ever having been best in previous iterations.  This seemed to be
suspicious to some.  Even after I posted analysis from Crafty where it does
_exactly_ the same thing, usually at least 2-3-4 times in _every_ game it plays.

I'm sure we'll here about it again.  Kasparov still brings it up.



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.