Computer Chess Club Archives


Search

Terms

Messages

Subject: Re:

Author: Moritz Berger

Date: 06:33:35 08/29/98

Go up one level in this thread


On August 29, 1998 at 08:30:29, blass uri wrote:

>If you use fixed positions then programmers may learn the weaknesses of their
>program in the fixed positions.
>
>Uri


But consider the enormous effort to just play comprehensive matches with the 10
"fixed" Nunn positions. If you just play 2 engines, chosing different positions
is feasible, but for e.g. 10 and more engines (remember, the intention is also
to specifically compare new releases with old versions, so 10 engines is an
understatement by far) round robin you need one set of positions to allow
meaningful comparison over time - as long as programmers don't start to "cook"
the positions. But primitive "cooking" (i.e. implanting hidden "books" from the
initial position) is easy to detect, as some programmers have learned in the
past from dubious results in the BT2630 test ...

Even with the fixed set of 10 positions (20 games per match), the task is almost
to hard to handle for one person/pair of autoplayers. That's why I asked for
support here.


Moritz



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.