Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: start your engines ...

Author: Tord Romstad

Date: 01:48:41 03/10/04

Go up one level in this thread


On March 09, 2004 at 22:37:10, Jonathan Kreuzer wrote:

>on an AMD 1.8ghz, time until mate in 9 is declared:
>
>Ruffian 1.01:     5.0 seconds
>Slow Chess 2.89b: 7.4 seconds
>Crafty :          ??? seconds (let it run for 2 minutes)
>
>I've noticed that Crafty quite often has trouble declaring mate, and I've
>wondered why (it doesn't hurt playing strength since Crafty is good at
>finding the best move quickly.)

I am not sure whether your guess is correct, but there is another possible
explanation.  My engine behaves similar to Crafty.  It is usually good
at finding the right move quickly, but often needs a very long time in
order to find a mate.  The current position is no exception.  It finds Rxg7+
instantly, but needs exactly 3 minutes before it announces "mate in at
most 110 moves".  Mate in 9 is announced after 4m28s.

The explanation for the long solution time is that my search is
intentionally lazy for the winning side.  When one side appears to
be winning, all search extensions for that side are severely limited,
and I search very few moves for the winning side in the qsearch (no
checks at all, for instance).  Instead, the search makes an effort to
find hidden resources for the losing side.

The idea is that it is more important to make sure there isn't a very deep
refutation of the (apparently) winning line than to look for an even more
crushing win.

In practise, this seems to work very well, even in tactical test suites.
I am sure my engine occasionally needs a handful more moves than necessary
to checkmate its opponent, but who cares?

Tord




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.