Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Endgame position revisited...

Author: Scott Gasch

Date: 13:08:51 04/24/02

Go up one level in this thread


On April 24, 2002 at 14:34:05, Dann Corbit wrote:

>On April 24, 2002 at 04:26:05, Scott Gasch wrote:
>
>>On April 23, 2002 at 12:40:09, Robert Hyatt wrote:
>>
>>>1b6/8/8/7p/6k1/6P1/8/6K1 w - - 0 1
>>
>>I take it you are not using EGTBs or node recognizers.  I turned off EGTB in my
>>engine but I cannot turn off the node recognizers... my output looks like this:
>
>What is a node recognizer?  Are you looking for repeats or cycles?

Ernst wrote a paper about node recognizers and I implemented it based on my
understanding of his paper.  Basically at all points in the tree (not just the
leaves) the search sends the board configuration to a recognizer function.  That
function, if the board is simple enough and it has the proper knowledge, sends
back a score and a type to the search.  For example one of my recognizers says
"if you have a rook pawn and the wrong color bishop and the enemy king controls
square X then it's a draw".

My main recognizer function calls out to the EGTB lookup to hit the tablebases
if there are less than 6 men on the board (and EGTBs are not disabled).  If
there's no help from them it calls the proper recognizer.  Most of the time
5-man EGTBs make recognizers useless... but sometimes I skip the EGTB probe
(don't want to probe TOO much) but I don't skip the recognizers because they are
high-speed (not relying on disk I/O).



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.