Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: How many Positions must a Program Check to complete a 9 Ply Search?

Author: William Bryant

Date: 13:56:13 07/30/99

Go up one level in this thread


On July 30, 1999 at 16:21:56, Terry Ripple wrote:

> A friend of mine who has a rating of USCF 2345 asked me that in order for  your
>favorite chess program to do a 9 ply search, how many moves or positions are
>there on the chess board that it would need to search in order to accomplish
>this goal?
> I hope i`am getting my point across to what i`am trying to ask for! It just
>seems that the numbers to do such a task would be in the billions for a complete
>search of 9 ply!
> Plus, they talk alot about how many Nodes per second a program can do and is
>there a formula to find out how many moves or positions your favorite program
>can do in relation to the KN/s?
>
>Thankyou in advance for any help in explaining this,
>Best regards,
>Terry

There isn't an absolute number of positions or nodes that need to be checked and
a simple response would be "enough to get the right move" :))))

Different programs will produce different node counts for the same search
depending upon the programs selectivity and the pruning of the search.

For example, I have done very little since adding attack tables to improve the
overall speed of my program in the last 3 to 4 months.  I'm still running at
about 50k NPS in the early middle game.  But, the time to get from 0 to 10 plies
has dropped significantly associated with more selectivity in the search.  In
other words the number of nodes that I have searched in order to complete 9
plies has dropped significantly.

The most recent change (that may still have a bug or three) was to add a
static exchanage evaluator and Q-search delta pruning.  By this, I mean that I
am dropping all captures that don't restore the material score back close to
alpha (ala Crafty -- Thanks Bob).  This has cut the time it takes me to reach
ply ten from the starting prosition of 1.e4 from 2:13 to 1:13 -- almost a 50%
time increase.  This means that I have cut the nodes I have searched by almost
50%.

I bet if you look all the top programs, the nodes to achieve ply 9 from any
given position will vary, possibly by a large number.

Hope this helps.

William
wbryant@ix.netcom.com



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.