Author: Severi Salminen
Date: 01:23:51 03/03/01
Go up one level in this thread
>>>>Yep, that's true. As Bob said it is very rare to have beta=INF. Actually >>>>Beta=INF can occur only in the very beginning of the search until we get our >>>>first real score (score<=INF-Ply). Remember also that you don't have to know >>>>anything else about the nullmove but whether its score is >=beta. So you can >>>>search with (-beta,-beta+1) >>>> >>>An idea I had, is that since about half the time we expect a move to fail low, >>>it doesn't make much sense to give the opponent an extra move and expect it to >>>fail high. >> >>Well, the point is: how do we know it _before_ trying the null move? Futility >>(pre) condition is one good indicator: if(depth==1&&material+MARGIN<=alpha). >> >We know from where in the search we are, e.g. if the move being considered is >off tje PV, we expect a fail high, and should try nullmove, but the response to >this move should fail low, and nullmove is a waste of time. Yes, now I understand, but if the move is off the PV why should we expect a fail high? It could be fail low move as well. And in normal implementations we already get these cases right: we don't try nullmove if we tried it at the previous ply (except Vincent). Severi
This page took 0.01 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.