Author: Robert Hyatt
Date: 09:14:39 01/14/02
Go up one level in this thread
On January 14, 2002 at 11:30:02, Severi Salminen wrote: >>How important is the Null Move Threat Extension in practical play? I would >>imagine that it solves some tactical problems faster, i.e. it scores better on >>WAC, but does it win games (or draw games that would otherwise have been lost)? > >Sorry my ignorance, but what is NMTE? If nullmove search returns something way >below alpha, you trigger an extension? > >Severi The idea is this. At some point in the tree, you will think some move is "best" (ie at a fail-high node). Before "believing this" you do a null-move search with the null-window lowered by some "delta" value (say one pawn). If the null-move search fails _low_, yet the one move you like fails high, then something interesting might be going on... such as this might be a horizon effect move that is holding off some serious threat. You go back and re-search the fail-high move, but one ply deeper, and you use the new score in place of the original. If it _still_ fails high, ok. If not, then you search for another move instead. If one fails high (which usually doesn't happen) then you again search it one ply deeper because you already know the null-move will fail low on the reduced window... This was written up in the ICCA 10 years or so ago... And it produces some interesting PVs, but it also will cost you at least a ply of search depth, possibly two...
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.