Author: Dan Newman
Date: 23:21:53 01/19/00
Go up one level in this thread
On January 19, 2000 at 20:37:58, Dann Corbit wrote: >Under exactly what conditions does NULL move degenerate? >Under exactly what conditions does NULL move remain valid? > >I see a lot of banter about the conditions of the test (when none were stated >except depth). What conditions matter and why? Null-move should work at any node where passing isn't the best move. There are two ways it can go wrong that I know of: 1) the node is one where passing is a good move (zugzwang) and 2) the search is too shallow. There are probably many positions where passing is the best move by a small margin. These are positions where the eval function has reached a "local" maximum and any move will spoil it. I don't think these probably cause much of a problem with null-move, but I don't really know. The ones that are real killers are where being able to pass saves material or the game. If null-move stays on for these positions it will make your program blind to how bad these positions are and so can blunder right into them. The endgame will of course have a much higher density of these sorts of positions since you have fewer pieces that you can just shuttle around to safe places. If the search is to a shallow depth, then the null-move search, being to a depth that is 2 or 3 ply less deep than the regular search, will likely miss important tactics, forcing moves and so forth and so will cut off things that shouldn't be cut off--it ends up causing the program to not see normally easy to see tactics. I think the idea is somewhat akin to the idea of "tactical sufficiency". If you search deep enough then it will see most of the short range stuff. -Dan.
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.