Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: How to Cheat in SSDF Competitions

Author: Dann Corbit

Date: 08:18:06 06/14/02

Go up one level in this thread


On June 13, 2002 at 23:52:27, Christophe Theron wrote:
[snip]
>You do not interpret correctly.
>
>I have never owned any Athlon chip, nor tested Chess Tiger on one.
>
>On only have computers with K6-2, P3 and P4 (I also have 386, Pentium and K5 but
>do not use them much anymore).
>
>Chess Tiger is also tested on DragonBall (68000 core).
>
>And to answer your question, the potential gain of knowing exactly on what
>processor the programs will be tested must be in the range +/- 1 or 2 elo
>points.
>
>What a bargain.

I do not know of *any* compiler which specifically targets an athlon core.
There are settings for PIII and above that Athlons will benefit from, but then
so will any of the other programs that use a PIII or better.  I have lots of
compilers (at least a dozen of them).  Not a single one has an "Athlon" swithch.
 Therefore, any special optimizations just for an athlon would represent a huge
amount of work.

Also, you probably cannot use any MMX instructions in a commercial chess
program, since people might buy it and run it on an old Pentium or something.
The cost of processing returns would make it a big hassle.

Quite frankly, I would be dissapointed at any chess program manufacturer who did
not target the higher chips in his compiler settings because most people will
have the higher chips (and yet the program will still run on slower chips).  If
anything, someone who targets a 386 is a nincompoop who is not benefiting his
customers.  And (as you say) targeting a CPU is not going to make any big
swings.

"An Honest Way to Cheat":
Implement learning.  All computer programs should do this anyway.  If they don't
it is a defect and the program deserves to be punished.

True cheating:
Custom books created just for the SSDF by playing huge numbers of games against
known opponents.  -- This would be a way to really distort the strength of
programs.  I would be very much surprised if it is allowed.  I imagine (without
knowing) that the programs used in the SSDF are simply bought off the shelf by
the testers and are not custom versions just for the SSDF contest.




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.