Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: The nature of the move from DTM EGTs

Author: GuyHaworth

Date: 15:21:18 06/16/03

Go up one level in this thread



Two contributions to ICGA_J cover this topic:

"Strategies for Constrained Optimisation", ICGA_J (March 2000)
"Depth by The Rule", ICGA_J v24.3 (September 2001) - a correction to the first.


There is another paper which depends on there being 'perfect information' in
EGTs.  This is an improvement on the 'first treatment' in the Computer Games
Workshop of the ICGA Computer_Olympiad_7 at Maastricht last June. ...

"Reference Fallible Endgame Play" ... to appear in ICGA_J (June 2003) I think.

This basically models an opponent, presumed fallible, and then picks tbe best
move to play against the opponent.  If the opponent looks fallible enough, there
is a non-zero chance that this is _not_ a metric-optimal move but one the
opponent is likely to struggle with, because there are lots of plausible but
non-optimal replies.


I won't anticipate the unpublished paper any further, except to say that
'metrics' define 'strategies of play' ... and 'strategies of play' can be
combined.  The first paper above explains this.

I do not now see a need to combine DTM data with DTZ data, although there is a
sensible way to do this.

g















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.