Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Can looking forward in hash table increase the number of visited nodes ?

Author: Ricardo Gibert

Date: 11:23:43 06/17/03

Go up one level in this thread


On June 17, 2003 at 08:06:00, Andrea Griffini wrote:

>On June 16, 2003 at 17:34:26, Ricardo Gibert wrote:
>
>>If you include leaf nodes, I can see how this can easily happen.
>
>I don't see how; can you elaborate ?
>
>>Try restricting its use to nodes further down the tree.
>>This can be generalized to restrict its use to nodes where
>>the remaining depth is greater than some constant parameter
>>"blah". The idea is the closer you are to the root, the greater
>>the payoff you get for a hash table hit. Also, a hit is more
>>likely, since for example, leaf nodes are more likely to be
>>completely new positions compared to interior nodes.
>
>My hash probing is very fast (or my move generation and move
>play/undo is very slow, depending on how you look at it) so
>probing all successors is not a problem.

After doing a little hunting, I found the following that may be useful to you:
http://citeseer.nj.nec.com/cache/papers/cs/2979/http:zSzzSztheory.lcs.mit.eduzSz~plaatzSzAAAI96-final.pdf/plaat96exploiting.pdf

If you want another format for document, look here:
http://citeseer.nj.nec.com/plaat96exploiting.html

>
>>Why don't you sort the moves for ETC? Just the first half dozen or so.
>
>Hmmm... I didn't realize that even if I can't find a beta
>cut may be I should push alpha up anyway if the stored
>position is trustable (i.e. the depth used to compute the
>stored value was sufficient). Is this correct or another of
>my misunderstandings about alpha-beta pruning ?

I think there may be a misunderstanding here. I wasn't suggesting you not look
at the remaining moves altogether, but to skip them in connection with ETC only.
Skipping altogether would be horrible.

The idea is probably not important anyway. Have a look at the above link. There
is plenty to do without worrying about my half-baked idea.

>
>Andrea



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.