Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Problem with draws by rep and hash table - solution?

Author: Brian McKinley

Date: 16:45:53 10/08/99

Go up one level in this thread


This discussion has given me some ideas.  I believe it may be possible to solve
this problem for draw by repetition without the sacrificing the advantages of
the hash table.

First lets break a draw by repetition into two categories:

1. The last repetition occurred before the start of our search.

2. The last repetition is part of our search.

In the first instance, we can safely store the hash value, because this position
will always return a draw value.

In the second instance, we need to discard all of the hash values we stored
after reaching the draw by repetition when we get back to the last repetition
for this position.  This could be implemented by switching to a new hash table
until we back up the search to the repeated position, at which point we can
switch back to the original hash table.  Of course to do this right, we would
need to keep a small stack of hash tables, and whether or not we need to start a
new hash when a draw by repetition is encountered then becomes dependent on the
node where we plan to discard the current hash table instead of the start of the
search.  The stack should remain small and these hash tables could be smaller
than our main hash table.

The 50 move rule has me stumped, but then it look like I am in good company. :)

Brian McKinley



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.