Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: using parameter depends on the opponent in the ssdf is unfair!!

Author: Tim Mirabile

Date: 14:56:38 03/20/99

Go up one level in this thread


On March 19, 1999 at 20:45:53, KarinsDad wrote:

>Issue #2: Program A has a feature that Program B also has, but it is turned on
>in the tournaments for Program A and not for Program B. This has validity if you
>are talking about features which are not defaults and which can also be turned
>off and on by the operators. If Program A has the feature by default and Program
>B does not, too bad.

I think this is the heart of the matter.  Programs are tested using the default
settings, correct?  And the default state of any program is "no knowledge of who
the opponent is" - regardless of whether it has a feature which could make use
of this knowledge.  By telling some programs who the opponent is, is the same as
changing its settings from the default.

So is this good/bad/fair/unfair?

That depends on what the true purpose of the SSDF list it.  If it is just a
sporting competition, with the idea of simply maximizing results against the
other programs on the list, then I don't see it as being terribly bad.  The only
thing a bit unfair here is that SSDF tests new programs against older ones
sometimes.  Older programs could easily have had this feature - it's not some
kind of new technology.  Perhaps this feature was never included in programs
until now because programmers had no reason to believe that SSDF testers would
actually use it.

Now if the true purpose of the SSDF is to provide consumers with a guide as to
the strength of various commercial programs for use in making purchasing
decisions, I don't see how using the feature would be helpful.  In fact, it
would skew the ratings even more than they are now in the direction of being
only accurate within the SSDF pool.

To take this to the extreme, someone could design a program which uses an
entirely different engine and book for each program on the SSDF.  If this
program is allowed to know which program it is playing each time, how could its
SSDF rating possibly be indicative of its strength against an unknown opponent
(i.e. the consumer).



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.