Computer chess
Andrew John Walker
as to whom can produce the fastest computer.To make computer chess
much more interesting, in computer chess tournaments all should be
required to operate on the same type of processor.As a result the
program that wins would be the one which has been programmed the
most effectively and has made the best use of available resources.
When programs of this nature are then used on faster computers,
they will be able to play stronger than brute force programs
such as Deep Thought.
Andrew Walker
-Broadband Network Maintenance
If the All Mighty Chess Computer is so very good with tactics, and only
brute force is needed to be the "god" of all chess then why do they have
an opening book from which to play the first moves?
I wonder what the ratings of the different chess computers and software
would be without their opening books? Or what would be their ratings
based only on their combination and tactic ability?
--Bruce
Jeff Lesser
Dr A. N. Walker
(Andrew Walker) [no relation, as far as I know, but how can I resist? :-)]
writes:
> To make computer chess
>much more interesting, in computer chess tournaments all should be
>required to operate on the same type of processor.
It's been done; Don Beal of QMW [Univ of London] ran a "Uniform
Platform" championship this summer. Unfortunately, the results were not
too promising for the Andrew Walkers of this world [but I'll get some of
the bugs out for next year].
>When programs of this nature are then used on faster computers,
>they will be able to play stronger than brute force programs
>such as Deep Thought.
This doesn't follow. Firstly, brute force often implies the
use of special-purpose hardware, which your hypothetical "program of
this nature" can't take advantage of. Secondly, the algorithms don't
scale; for example, faster computers often achieve their speed through
multi-processing, through subtle uses of caching and other similar
features, and so on. To take advantage, you really need to parallelise
your program; but how this should be done depends strongly on all
sorts of features of the machine architecture. Thirdly, why should we
assume that the winning "uniform platform" program will itself not be
brute force?
Even if you simply take a program from one computer and run it
on another reasonably similar computer, to take best advantage you
probably need to tune all sorts of parameters (such as the sizes of
various tables). The existing brute-force programmers have lots of
experience of this process, which may well keep them ahead of the game.
--
Andrew Walker, Maths Dept., Nott'm Univ., UK.
a...@maths.nott.ac.uk
[There's yet another of us who appears on "rec.sport.cricket".]