Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Question for programmers about nonsense PV's

Author: Robert Hyatt

Date: 18:28:20 06/11/00

Go up one level in this thread


On June 11, 2000 at 20:55:15, Robin Smith wrote:

>I have heard it said before in this newsgroup, that the reason PV's that are
>displayed sometimes have nonesense moves, moves that give up material for no
>reason for example (I have even seen ILLEGAL moves displayed in PV's!), is that
>hash tables allow a program to know WHAT the evaluation of a position is but not
>HOW (true PV) it came about.  Is there ANY way to have programs only show those
>moves that can be KNOWN to be part of the PV that leads to the eval being
>displayed?


All programs don't do this.  IE in Crafty, I actually "back up" the PV along
with the score, so that my PVs don't ever contain nonsense moves.  I have tried
extracting the PV from the hash table, but it can cause the problem you are
describing...




>
>The reason I ask is that I think the present way of displaying PV's is
>misleading to non-programs such as myself.  One often hears comments like "the
>computer plays the right move, but for the wrong reason".  Probably many times
>it is for the RIGHT reason, but the PV that was displayed does not display the
>reason.  Anyway, I think NO information is usually better than WRONG
>information.  I would much rather see a short but accurate PV than a long and
>misleading one.  I mean to say, ILLEGAL PV's??  Come on.  It must be possible to
>do SOMETHING better than this!
>
>Programmers out there, what are your thoughts?  Also, any estimates as to what
>percentage of displayed PV's DO and what percentage DON'T display the correct
>moves (the line that leads to the eval being displayed)?
>
>Robin Smith



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.