Author: Maurizio Monge
Date: 16:45:18 10/26/05
Go up one level in this thread
On October 26, 2005 at 14:15:08, Tord Romstad wrote:
>On October 26, 2005 at 13:10:45, Maurizio Monge wrote:
>
>>Hi,
>>I have found that 2 engines ("TRACE"
>>and "Petir") use this extension.
>
>SmarThink and Gothmog use them, too. The extension was
>introduced by Sergei Markoff (author of SmarThink), and was
>apparently inspired by some idea in an old paper of Botvinnik.
>Sergei modestly named the extension "pseudo-Botvinnik
>extension". I implemented them in Gothmog, noticed a
>small, but clear improvement, and introduced the more
>attractive name "Botvinnik-Markoff extension". :-)
>
>>Does anyone know what it is?
>
>Extending when the null move is refuted by the same move,
>or by the capture of the same piece, two times in a row
>(i.e. at ply N and at ply N+2). The point is to avoid horizon
>effect problems, for instance when there is an unavoidable
>mate or an unstoppable passed pawn, but the defending
>side can postpone disaster for a few moves by delivering
>checks.
>
>Tord
Thanks for your detailed anwer, Tord.
So, if i understood correctly, the extension is activated
when the null move fails because of a strong reply and the
reply is still avalable for the other player after 2 plys.
It means that, to be able to do this, it is no more enough
to do the null move search with a zero window (-beta,-beta+1),
like my experimental program do at the moment (Fruit 2.1
does this too), because this way you could miss the strong
reply...
I guess if it is worth, maybe later a test will persuade me
that it is :)
Thanks!
Maurizio Monge
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.