Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Cray Blity 1981?

Author: Robert Hyatt

Date: 11:58:51 06/02/00

Go up one level in this thread


On June 02, 2000 at 14:13:49, Joshua Lee wrote:

>
>>
>>I was present at that tournament.  It was a wild affair (that game) with both
>>sides failing high, failing low, etc.  Chess 4.x was searching about 2500 nodes
>>per second, Belle was searching about 5,000 nodes per second.
>
>As i go through these games (usually the winner of the ACM that particular year)
>i've noticed that the comp's miss things is this due to the search depth or is
>this an eval problem?  for example i am looking at Cray Blitz - Belle  12th ACM
>1981 and it looks as though 28.QxNb6 is a mistake where Bxh6 may hold a draw.
>I am not sure weather or not i can find a win for CB but from what i have seen
>games from 1983 and on are going to be a bit more difficult due to the greater
>search depths and speed. Except for that Bebe-Hitech Game those two then don't
>compare to now even Hitech improved to 200Knps in 1995 but i am not sure if bebe
>stayed the same or just stopped competing. Lastly i did notice that the pc
>programs in 1994 against Deep Thought were really slow, what can you tell me
>about these events or where might i find detail info online? thanks again
>Dr. Hyatt



Bxh6 has been a test suite position for many years.  Here is what happened to us
that year.  Right at the last minute, our primary machine at Cray Research
became unavailable due to a complete moving of their corporate research lab to
a new facility.  They made arrangements for us to use a pretty old cray at
the University of Minnesota.  The only problem was that the Minnesota machine
didn't have the "interactive access facility" hardware, which meant it could
only be used via batch processing.  To get CB to make a move, I edited a batch
file which had all of the current game's moves included.  I would add the new
move, then submit the job, which went thru a queueing front-end machine, and
on to the cray.  To be safe, I had to set the search time per move to about 1
minute, rather than 3, even though we were playing 40/2hr time controls.  I
could not use any pondering or anything.

On that move, Qxb6, we went 1 ply too shallow to see the forced draw by Bxh6.
Had we been "interactive" so that we would have been using normal time controls,
we would have seen that even back then.  Unfortunately, we didn't.  There were
other such things that happened to us over the years, so it was just "part of
the normal course of unexpected events."


PC programs were _always_ slow.  Even on occasions when we were not as "smart"
as our opponents, we could usually out-search them anyway and _still_ win.  Of
course, we worked on those "holes", but as we found out, speed covers up for a
lot of things.



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.