Author: Christophe Theron
Date: 07:35:47 07/22/99
Go up one level in this thread
On July 21, 1999 at 19:04:52, Vincent Lejeune wrote:
>I've found the original post
>
>I've verified the position but Fritz find the right move (h4 !) without problem,
>so I doubt that's a nullmover killer position, or Fritz have special algorithm
>for this kind of position ...
I still think it's a null move problem.
Null move pruning depends on the values of alpha and beta, and also on the move
ordering. For example, no null move cutoff can happen just after the first move
you look at in the root position. If by luck you look at the key move first, you
will solve the problem even if the other side is in zugzwang at this point.
The null movers can fail *randomly* on some zugzwang positions. I mean they will
usually fail to solve them, but in some occasions, and by luck, they will still
be able to solve them.
A number of factors can influence this:
* the evaluation (2 different programs have different evaluations, and thus the
values of alpha and beta will be different, and the search behaviour too)
* the move ordering
* the content or size of the hash table
I understand that some users will find that this is a nasty behaviour of their
favorite chess program, but that's the price to pay to achieve _in the general
case_ very good performances.
Christophe
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.