Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: chain of responsibility

Author: Robert Hyatt

Date: 13:42:06 04/29/00

Go up one level in this thread


On April 29, 2000 at 16:17:47, Dave Gomboc wrote:

>On April 29, 2000 at 12:22:35, David Eppstein wrote:
>
>>On April 29, 2000 at 09:47:23, Robert Hyatt wrote:
>>>There is one small problem here, but it is very serious.  You can't really
>>>afford to turn a module on or off.  Doing so causes discontinuities in the
>>>evaluation and can cause really bizarre moves to be played.
>>
>>If the added discontinuity happens at a point where the eval is discontinuous
>>anyway (as in Bruce's program, it only happens when the material changes) maybe
>>the edge effects are not so severe?
>
>At first read, Bob's post seemed to be tangential to what I meant.  I was
>thinking about something like:
>
>1) endgame db probe
>2) endgame "subtraction" db probe
>3) hash table lookup
>4) recursive search/eval
>
>If a step returns a successful answer, you're done.  Different chains can be
>prepared at either compile-time or run-time, as needed.  Hmm, I think my example
>here is probably not very good, but I've got to get back to work. :)
>
>Dave


I think most do that already.  IE I first probe the hash table, then I probe
the EGTB, then I call eval which first checks for clear draws.  If it finds
one it returns.  If it finds no pawns it does a special mate evaluation.  If
it finds pawns, it does a normal lazy evaluation.

That is ok, as you first try the most accurate thing you have, and if that
fails, you try something less accurate, and so own, until you finally try
your heuristical evauation.  But inside the evaluation, if you have
discontinuous evaluation terms, they cause problems around the boundary where
they turn on or off.

There are plenty of similar mathematical formulas that are discontinuous at a
specific point..  IE as you approach the point, starting from 0 on the x axis,
the y value gets smaller and smaller, until at the point of discontinuity, it
reaches -infinity.  Step one unit further and you start off at +infinity and
the Y value starts to drop. back toward zero.  Right around that point, things
are _very_ ugly.



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.