Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Tiger SMP: why ?

Author: Robert Hyatt

Date: 09:38:18 05/02/01

Go up one level in this thread


On May 01, 2001 at 09:40:38, James T. Walker wrote:

>
>Hello Tony,
>There was no requirement for Tiger to be SMP.  Since Christophe stated he
>already had an SMP version and could make it ready for competition in a few days
>that was not the reason Tiger was denied a chance to play.  It seems to me that
>Tiger was systematically eliminated by the people running the tournament for
>some other unknown reason.

There is nothing to support that conclusion.  There is _no_ public evidence
of an SMP tiger.  The author has gone on record multiple times saying he was
not interested in such.  Since the people organizing the tournament were not
mind readers, it isn't surprising that Tiger wasn't asked to participate.

I think it much more serious that Ferret was excluded.  It _has_ been SMP for
a long time.  Mine was the first that I know of.  Bruce was next.  The others
were started well after those two were already playing games on ICC reliably.



> We are all only left to speculate as to what the
>real reason was.  Personally I believe Deep Shredder should have been the one to
>play without a qualification tourney but if a tourney had to be held then Tiger
>should have been permitted to join as an SMP version.  Since 3 programs cannot
>play at the same time, there was no real reason that Tiger could not have
>entered after a few games already played between Fritz/Junior.  Another point is
>why are they playing on  2 processors if the final will be on 8 processors?
>Seems like the sponsor should have provided two 8 processor machines  to run the
>tourney on since they will play entirely different on 8 vs 2.  Since Kramnik is
>demanding a copy of the winner to practice against and find it's weakness I'm
>surprised that he did not also demand an 8 processor machine to run it on.  This
>entire thing is as phoney as a 3 dollar bill.

I don't think it is "Phony".  It is just scientifically embarassing.  The goal
of a scientist is to first pose a question that needs to be answered, then
construct an experimental setup that will answer that question.  This tournament
did _neither_.

The right question (assuming that Shredder was not going to be chosen due to its
recent ICCA tournament victories) was "Which program will do best against humans
using an 8-way SMP box?"  The mistakes after deciding on the above question are
numerous.  1.  They didn't use an 8-way box for the competition.  A dual does
_not_ show how a program will do on an 8-way box.  There are good and bad ways
to do parallel search.  Either will do ok on a 2-way box.  But not on an 8-way
box.  They played computers vs computers to choose the program.  In that case
the ICCA event was more reliable since it had more opponents for each program
to play (rather than just 1 in the match actually done).

And it excluded other SMP-ready programs that might do as well or better than
Fritz against humans.  I can think of two programs that play human GM players
_all_ the time on ICC.

In short, it was flawed from the beginning, poorly set up, and was no better
than simply flipping a coin between Fritz and Junior...








>Just my humble opinion.
>Jim



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.