Author: Robert Hyatt
Date: 11:43:38 11/13/03
Go up one level in this thread
On November 13, 2003 at 11:07:09, Daniel Clausen wrote: >On November 13, 2003 at 10:38:35, Robert Hyatt wrote: > >[snip] > >>>Imagine a position where white can exchange rooks and reach the famous FINE-70 >>>position. With HTs it's quickly possible to see the win. I'd like me engine to >>>see the win before the exchange though, not afterwards. (since then, I maybe >>>never reach the position in the first place) >>> >>>Yeah, I know these are just words. I mean, at least John tried something _for >>>real_ and it worked. :) But maybe one day... :) >>> >>>Sargon >> >> >>This just opens a huge can of worms that has been discussed thousands of >>times in the past. Here is the question: >> >>"what makes it reasonable to search one root move deeper than another?" > >Not sure why you're talking about root-moves specifically here. My suggestion is >meant for everywhere in the tree. (which I'm sure will explode like hell, but >hey, lemme try it out too! =) I think John proposed and tested this as a fix for searching root moves only... > > >>sub-questions include "if you search two moves to different depths, how can >>you possibly compare the scores?" Answer: "you can't." Of course, if you >>search two moves to different depths and the deeper search proves a win, >>then you can safely take that even if the other move also wins quicker if you >>search it to the same depth. But if it isn't winning, then comparing the two >>scores is _very_ risky. > >Well, since we do that (compare scores of adjacent nodes with different depths) >all the time, why should this be a killer for the idea above but not for all >extensions? Sounds unfair to me! =) > >Sargon
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.