Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: EGT_generation observations

Author: Heiner Marxen

Date: 08:16:01 06/01/02

Go up one level in this thread


On June 01, 2002 at 10:33:39, GuyHaworth wrote:

>
>Typically, in the generation of an EGT, the 'draws' are those 'unbroken'
>positions that have not been resolved as wins for one side or the other.
>
>'Draws' are not given a depth.
>
>However, 'draws' can be partitioned into 'draws requiring conversion' and 'the
>rest'.  Trivially, many draws require the material-down 'Black' to capture a
>White piece immediately .. or some similar tactic.
>
>Such draws can be given a 'DTC' or 'DTZ' depth by a technique similar to the
>technique used to find wins.  This has not so far been done.
>
>Maybe this is relevant.

The only application I can think of is to try to assign some degree of
"difficulty" or "non-trivialness" to draws.  But that would only help
for "swindles", i.e. when one tries to give the opponent opportunities
to do a mistake.

Well, humans _do_ think that way, sometimes, so maybe there is something
to gain here.  I have no idea whether it could turn out to be important.

Wait a second... it just occurs to me, that I could even use it in Chest...
Some days ago Dann sent a feature request for Chest to me, asking to
implement a draw recognition: instead of searching for a mate (win)
searching for (and proving) a draw.  My answer basically was "what is
the depth of a draw?".  What you describe may be (part of) the solution
for this.

Thanks!

>g

Cheers,
Heiner



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.