Xref: utzoo rec.games.chess:3754 gnu.chess:88 comp.ai:5451
Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!cs.utexas.edu!uunet!clyde.concordia.ca!mcgill-vision!quiche!utility
From: utility@quiche.cs.mcgill.ca (Ronald BODKIN)
Newsgroups: rec.games.chess,gnu.chess,comp.ai
Subject: Re: A Study by Reti
Keywords: chess programming
	Fidelity probably uses one of two possible "tricks" in use here:
Message-ID: <1938@quiche.cs.mcgill.ca>
Date: 9 Jan 90 19:22:12 GMT
References: <5785@rice-chex.ai.mit.edu>
Followup-To: rec.games.chess
Organization: SOCS, McGill University, Montreal, Canada
Lines: 11
Reply-To:

(I'd bet on this -- it has a fairly good idea that one should chase
enemy pawns and/or protect your own in king-pawns endgames)
2) an outright lookup table at some point (very dubious, given the
-0.25 evaluation, where a lookup table would give 0.00 -- I'm curioushow
it could produce such an evaluation -- probably they don't have it
instructed that king vs. king is a draw (so one king is considered "safer"
or "more mobile" by the program).  Note that people have constructed
such tables for up to 6 pieces, and I'm sure KP vs. KP has been constructed.
	I've actually written a chess program, so I'll feed your problem
to it and report the results.
			Ron