Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Fail-low pruning

Author: Robert Hyatt

Date: 13:45:41 01/17/06

Go up one level in this thread


On January 17, 2006 at 16:27:42, Dann Corbit wrote:

>On January 17, 2006 at 16:23:53, Tommi Rimpiläinen wrote:
>
>>Hello!
>>
>>I suppose all the major chess programs use Tord Romstads Fail-low pruning now a
>>days. Fruit uses it and it is likely that Rybka uses it also. It would be
>>interesting to know, what the idea is behind this method of pruning.
>
>If it looks like crap, don't search as deep.
>
>>How sound
>>is this idea theoretically?
>
>All pruning beyond Alpha-Beta is unsound (IOW, it could cause you to miss
>somthing and get the wrong answer).  However, empirically, it often works quite
>well in practice (e.g. null move pruning is an example of theoretically unsound
>pruning that is so successful practically everyone uses it.  And those that do
>not use it either use some related idea or lose all of their chess games).


Note that null-move is not _that_ "unsound".  The "null-move observation" is
actually pretty well-founded in the game of chess.  That is, if I can make a
null move (pass) and you get two consecutive moves and still can't kill me, your
position really sucks...  That is not quite as unsound as just saying "I am not
going to search this move at all, or I am not going to search this move as
deeply."  Both of which can introduce more errors than null-move easily.  NM's
main adversary is zugzwang...



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.