Author: Robert Hyatt
Date: 11:36:53 01/09/98
Go up one level in this thread
On January 09, 1998 at 11:26:20, Bas Hamstra wrote: >Bob and Don thanks. > >I hope you don't mind I have more to ask about singular extensions. > >So the basic idea is to grant an extension if one move is much better >than the rest, to follow forced lines, much like it is done >with checks. But checks are *absolutely* forcing, while singular moves >are not.... > >Example: you are at D=9 and find a singular move. Any other move loses a >pawn, so extend. In fact you discard other moves. But maybe that is NOT >correct! Maybe if you would search to D=13 you would find that one of >the moves you discarded wins a piece!! So how reliable are the forced >lines you build? A specific question: > >Suppose a 9 ply search leads to a 17 ply forcing line with score >+3 pawns. Is it a garantueed 17 ply score then? No 17 ply brute force >search could ever prove the score being less? > >That's what I keep thinking about. And I have the feeling the answer is >NO, a 17 ply BF search could still prove it being false (very much >unlike chess extensions I would think), because moves have been >discarded on the basis of a 9 ply search. Of course it could still >"work", but as no more than an inexact selection heuristic. > >Comments? > Note that this only extends an extra ply here and there. *no* forward pruning of any kind. So the thing you see is totally forced along that line, just like checking lines are totally forced... not because they are checking moves, but because you extended and confirmed what's going to happen. And you extend because you know there aren't so many checking moves that extending them all will kill you... So it is no less accurate than "get of out check".. >The second question is: has it been tried to steer the game into a >direction where the opponent has lesser choices? > > > >Regards, > >Bas Hamstra.
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.