Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: R=3 question

Author: Anthony Cozzie

Date: 05:46:36 06/03/04

Go up one level in this thread


On June 03, 2004 at 07:58:52, Tord Romstad wrote:

>On June 03, 2004 at 07:37:30, Anthony Cozzie wrote:
>
>>On June 03, 2004 at 03:21:07, Daniel Shawul wrote:
>>
>>>Hi
>>>
>>>I used to use adaptive null move pruning.
>>>But now since i added checks only in the first ply of quiescence search(horizon)
>>>I think it is time to use R=3 everywhere. Is it safe?
>>>Do i need to do other safe guard searches like verified null move
>>>pruning?
>>>
>>>best
>>>daniel
>>
>>I use R=3 everywhere (checks in q-search).
>
>Even in the endgame?  I never had any success with R>2 in the endgame.

Sorry, R=2 if there are only a few pieces left.

>>It is not a pure win though: you see more in some areas and less in others.
>
>You can try to compensate by using a dynamic reduction factor.  I use R=3
>most of the time, but depending on the position it also happens that I
>use R=1, 2 or 4.  The main factors are the amount of material left on the
>board, the king safety for the side to move, and the passed pawn eval for
>the side not to move.  The idea is to avoid using a too big reduction factor
>in positions where the risk of horizon effect problems is big.
>
>My experience is that a dynamic reduction factor works better than all the
>more common alternatives (pure R=2, pure R=3, verified null move pruning,
>and adaptive null move).

Yes, you have stated your heretical ideas here before :)

I think that R=3 is more stressful on the evaluation function than adaptive.  It
requires a better kingsafety & passer eval.

anthony



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.