Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Bounds... when is an exact score an EXACT score?

Author: Robert Hyatt

Date: 06:58:45 08/01/03

Go up one level in this thread


On August 01, 2003 at 02:33:13, Ross Boyd wrote:

>Hi all,
>
>I want to get a clear definition of hash table bounds...
>
>So, this is what I understand...
>
>When we get a score during search, one of 3 things can happen:
>
>1. The score is >= beta... so we store this as a LOW_BOUND.
>2. The score is <= alpha... so we store this as a HIGH_BOUND
>3. Or its between alpha and beta... so its an EXACT score.
>
>So the definition of an EXACT score is anything we find between alpha and
>beta...
>
>So far so good....
>
>Ok, what about an EGTB probe? A successful EGTB probe returns a true score
>whether it be a draw or the number of moves to mate.

You are mixing apples and oranges.  Suppose you get a hash hit with an
EXACT indicator, but it is outside the alpha;/beta window.  What would
you store for such a position later?  Answer = EXACT.  Because it is an
exact score.  An EGTB hit is the same.  It is not the result of an alpha/
beta search below this position with the current bounds in force, it is
the result of a _complete_ search from this position.

Complete searches always produce EXACT scores.  Searches with an alpha/beta
bound to speed up the search only produces an EXACT score when the score lies
within the alpha/beta window.





>
>By applying the 3 rules above I SHOULD store the EGTB score as a LOW, HIGH or an
>EXACT bound...  Is that correct?

No.  EGTB hits are _always_ EXACT.


>
>Why not store it as an EXACT score?  (Its as exact as you're going to get...)
>Will the ab search break in some subtle way if I do this?
>
>I'd like to understand this once and for all....
>
>Thanks for your comments!!
>
>Ross



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.