Computer Chess Club Archives




Subject: Re: Rybka uses PVS and not MTD(f). Its no Fruit-Clone

Author: Ryan B.

Date: 02:19:49 12/12/05

Go up one level in this thread

On December 12, 2005 at 04:58:52, Chrilly Donninger wrote:

>I did not follow the CCC-discussion the last months/years. So please do not
>"kill" me if I am telling old stuff. I got questions about Rybka from the
>Hydra-sponsor so I read yesterday evening some of IM Vasik Rajlichs very
>interesting postings in the archive.
>Vasik mentions several time MTD(f) and his experience with it. But when playing
>with Rybka, it looks like a normal PVS-search. I looked therefore also in the
>Rybka-Code. What I have seen so far is typicall PVS-Code.
>From this first look I can say that Rybka is not at all a Fruit clone.

You could look at Fruits source or just read the readme files :)

Search algorithm


The main search algorithm is a classical PVS with iterative deepening.

Search enhancements such as a transposition table and null-move

pruning are also used (see below).

A few details in the PVS implementation are not-so-standard and are

there to supposedly enhance the stability of the search (like reducing

the consequences of search inconsistencies).  For example the

re-search window after a scout fail high of score "value" (with value

> alpha) is [alpha,beta], not [value,beta].  As another example, I

only allow null move when the static evaluation fails high

(i.e. eval() >= beta).  Whether these features improve the strength of

the engine is an open question.


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.