Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Interesting mate test for hashing

Author: Dave Gomboc

Date: 00:34:17 09/11/99

Go up one level in this thread


On September 11, 1999 at 02:28:00, Ed Schröder wrote:

>>There are only three choices here.
>>
>>(1) best <= alpha.  We don't know anything about any of the moves.  Yet I
>>believe Ed (and I thought you as well) said that if you take this 'best'
>>move and store it in the hash, it works well.  In my test, it didn't.
>>
>>(2) alpha < best < beta.  This is a move with an _exact_ score that is
>>correct.  Also known as a PV-candidate move normally.
>>
>>(3) alpha < beta < best.  This is a fail high move.  It is better than the
>>best that is allowable.  It is _not_ the best move at this level however,
>>it is just good enough to produce a score >= beta and terminate the search early.
>>This is the case that feeds (1) above at the previous ply.  If we try
>>_another_move here we might get an even _bigger_ score here, and back up an even worse
>>score to the previous ply.  But alpha/beta makes us stop _now_.
>
>I like to try this. One question: what is "best"?
>
>the score of the node?
>
>the best score sofar on this ply?

"best" is initialized to -oo before the subtrees are searched.  At any given
time, "best" is the best backed-up score or bound from the subtrees that have
been searched.

>Ed

P.S.  Why do I always see INF or INFINITY in people's code?
      I always use "oo". :-)

Dave



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.