Author: Vasik Rajlich
Date: 02:27:41 02/19/04
Go up one level in this thread
On February 18, 2004 at 08:29:15, Andrew Wagner wrote: >Yes, I'm STILL having trouble with this. I'm attempting to emulate Bruce >Moreland's algorithm at http://www.seanet.com/~brucemo/topics/pv.htm and having >trouble. Can someone explain to me exactly what's going on there as far as >updating the PV? Is it assumed that MakeNextMove() and UnMakeMove() update the >line, so that when it hits a fail high it knows what moves were made to get >there? I just don't understand his explanation. I've tried numerous ways of >doing it, and can't get it to work. But I know I'm getting the root move >correctly...it's actually able to play a game now (though not too well). Any >help would be appreciated. Thanks! IMO the PV is only useful for debugging and cosmetic purposes. The move ordering should come from the hash table. In cases where a PV move is not the same as the hash table move, the PV move will give you the information you need to understand the score which your engine reports, but the hash table move will always be better. (This assumes that you only store fail-high moves in the hash table.) For example, imagine some variation which turns out to be the PV in which a position P is reached with depth D, and fails high with move M1. Later in the search, the same position may be reached with depth D+2 (for example there was a pair of shuffling moves in the PV line), and now M1 fails low while a new move, M2, fails high. M1 will be the move in the PV, and will explain the score your engine reports. M2, however, is clearly the better move. Vas
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.