Author: Alessandro Damiani
Date: 23:10:04 03/11/99
Go up one level in this thread
On March 09, 1999 at 06:28:42, Steve Maughan wrote: >On March 09, 1999 at 03:46:16, Alessandro Damiani wrote: > >Allessandro > >Thanks for the feedback - I'm afraid you lost me with some of it! > >>Hi! >> >>I have changed from Alpha-Beta to NegaScout because of Singular Extensions. The >>big advantage of NegaScout is that at the root of the search 'tree' you know >>which move will be the PV-move: > >I don't understand!? > >Surely with Alpha-Beta and iterative deepening you also know which move will be >the PV move at any point? > What I mean is: at the root you have n moves. At iteration i you have searched k (k<n) moves. So you have a best move and a pv. Now you test with null-window the k+1th move. If you get a fail high you know(!) that the path starting with this move will be the new pv (assuming no search anomalies). When researching this move I activate Singular Extensions. The consequence is that if after the research the move is still better than the current best move, all singular moves on the pv have been searched deeper. So the research is more exact but costs more. By using any kind of extensions or forward pruning it can happen that the test tells you, the current move is better than the current best one but the research tells you the opposite. This can happen with Singular Extensions. >>if the null-window test tells you that the >>current move is better than the best move it will be researched and get the best >>move. So NegaScout gives you more information about the pv than Alpha-Beta. If >>you use heuristics that change the length of the paths of the 'tree' then you >>will notice search anomalies (contradictions between a null-window test and a >>research). To avoid this I have removed the fail-soft part of NegaScout, >>accepting the consequences. >> >>In Fortress I only use Singular Extensions at pv-nodes. Using NegaScout the >>program accepts a pv only if all singular moves on it are searched deeper and is >>still best. Of course the cost is high! But I hope that a better forward pruning >>will change that (the current forward pruning is too risky). >> > >What do you mean by "accepts a pv only if all singular moves on it are searched >deeper and is still best"? How are you defining what is and is not a singular >move? First question is answered above. I simply use the definition of singularity the authors of Singular Extensions published: a move is singular if its score is at least S higher than the scores of the siblings, S being an integer. > >>Alessandro > >Regards > >Steve Maughan
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.