Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: One-reply extension howto

Author: Sune Fischer

Date: 05:25:03 03/05/03

Go up one level in this thread


On March 05, 2003 at 07:37:26, Georg v. Zimmermann wrote:

>On March 04, 2003 at 07:12:21, Sune Fischer wrote:
>
>>On March 04, 2003 at 05:45:53, Georg v. Zimmermann wrote:
>>
>>>>While on the subject, how does one detect singular moves?
>>>>
>>>>-S.
>>>
>>>Shallower search for all possible moves on that node (Overhead). If all but 1
>>>move fall under a certain thereshold, that 1 is singular.
>>>
>>>Georg
>>
>>I just don't see how that works with alpha-beta. Is it a low-level minimax?
>>
>>-S.
>
>Why not ? Maybe I am missing smth.
>
>After you get the best move or fail high at a node you search all other moves
>with a lower depth and a window at beta- margin (like 1.2 pawns).
>If none of those searches fail high, they are all worse than the found move =>
>you research the best move with depth+singular_extension.
>
>The only situation where this fails is when you failed high on (i.e.) the 2nd
>best move and during the depth-reduced search find the "real best move", which
>of course fails high. Now you skip the singular extension even if the real best
>move was actually singular, because it is much better than the first best move
>found. But that case can be ignored since in such a situation the side to move
>is already doing so well (2 ways to fail high, 1 of them even much better) it is
>probably not a relevant node.

I don't see any reason to extend a move I fail high on, unless there is a threat
further down. In that case I would prefer a heuristic that extends on threats
directly.
Extending singular moves like PxQ where there is no threat involved is a waste
of time as far as I can tell.

But thanks for your technical explanation :)

-S.

>Georg



This page took 0.01 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.