Author: Robert Hyatt
Date: 07:10:25 08/18/01
Go up one level in this thread
On August 18, 2001 at 09:23:13, Alvaro Jose Povoa Cardoso wrote: >I think the problem might be related with extensions. >I use fractional extensions that may be diferent for each move at this node. >At the top of the search I don't have all the information I need to compute the >extension for the hash move. >For example, I use the number of moves at this node in a formula to compute the >extensions for every move I search, since I don't have that at the top of the >search I cannot calculate the extension value for the hash move. >I think I have to rethink the extensions in my program. > >By the way, can some one explain how does it work in crafty. >If I'm not mistaken it only uses a single bit in the hash table to tell if we >should extend or not. The question is how much I should extend? A full ply? >I so that is not consistent because Crafty uses fractional extensions, right? > >Thank you all, >Alvaro Cardoso In crafty, the "extend" bit is for threat extensions dealing with mate. I use the approach that if the null-move search returns "mated" then this is a special-case position such that if I do nothing I get mated, if I do something I don't. That means there is a strong mate threat on the board. When I store the entry in the hash table, I store a single bit as well, so that when I do the search here, I will remember to do the mate threat extension as well. Note that the hash table can also say "no need to try a null-move search here" which would turn off the mate threat detection without this bit. This "flag" is _only_ for the mate threat extension, so the fractional ply extensions don't cause problems. Only one is triggered by this flag... the mate threat extension.
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.