Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Pondering accuracy?

Author: Robert Hyatt

Date: 05:22:51 11/13/98

Go up one level in this thread


On November 12, 1998 at 20:26:43, Peter Kappler wrote:

>
>Fellow chess programmers:
>
>My chess programming project for this weekend is to add pondering to my program
>and so I'm curious - do any of you keep statistics for how often your program
>ponders the correct move?  I recall that Bob Hyatt once said that Crafty ponders
>the correct move ~50% of the time.  (Bob - please correct me if I'm wrong.)


In general, that's about the worst I see.  Against other programs it is way
higher.  Against GM players it is way higher (in the last GM match I had
Crafty analyze online it predicted properly well over 90% of the time for
both players...)

In some positions it will do poorly, such as during the opening when there are
several ways to develop pieces and it is a matter of taste which goes first and
to which square.  In tactical positions the opponent had better match it most
of the time.  In endgames, anything can happen.

We had this discussion a year or two ago... to answer the question I added the
"p:31" type output to crafty's whispers on the chess servers... that is the
number of correct predictions...  I hardly ever see it under 50%, and most of
the time it is much higher...  but anyone can watch to see...




>
>It would be particularly interesting to compare the %correct in games versus
>computers and games versus humans.  It should be significantly higher in comp
>vs. comp games...
>
>It would be neat to compare this percentage for various programs, and see how
>these data correlate with the playing strength.  (Only meaningful if the data
>were computed from games played against reasonably strong opponents, I
>suppose...)
>
>
>--Peter



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.