Author: Scott Gasch
Date: 12:23:37 09/12/04
Go up one level in this thread
On September 12, 2004 at 14:48:32, Dieter Buerssner wrote: >Maybe, instead of making the extension decision dependent on alpha, make it >dependent on the material score or static eval or perhaps even some exact >(= open window) shallow depth [q]search score. If I am looking lost already, >don't extend single move replies to check (and perhaps, if I am winning already, >don't extend the check). So instead of saying "we know side on move has something better elsewhere therefore he is down here" you are saying "we know the material or score at the root is way better than the material or score here... so I assume we are down here"? The only downside of this that I can see if if every move from the root loses you will not extend anywhere, right? I already do something like the above for recaptures: I only consider them near the horizon to qsearch and I only extend them if the material balance of the node looks like it's close to the balance at the root of the search. I guess the same idea could be applied to a lot of extensions... >Deciding about the extension upon alpha (or beta) will add inconsistencies to >the search, that will also depend on hashing. > >BTW. I did exactly that change: window dependent extension decision to material >balance based extension decision. Did not see much of a difference, however.
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.