Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Question For Robert Hyatt

Author: Robert Hyatt

Date: 07:04:13 02/03/99

Go up one level in this thread


On February 02, 1999 at 23:32:56, Melvin S. Schwartz wrote:

>You stated that Selective Search looks deeper but with intrinsic error. Then you
>said Brute Force takes longer but will get to the same depth without the errors.
>Now, how could you say Selective Search looks deeper and then say Brute Force
>takes longer but gets to the same depth?


take a game where each side has exactly 10 moves at each 'node' in the tree.
If the 'brute force' program searches to depth=5, it will search 10^5, or
100,000 nodes.  If that is a limiting factor (it usually is, in that most
programs search a reasonably similar number of nodes per second if they are
pretty equal in skill) then you are stuck with 5 plies until you can search
faster.

The selective program might choose to only search 5 of the 10 branches from
each node (hence it 'selects' which 5 to follow, commonly called 'forward
pruning').  And as a result, it can search to depth = log5(100000).  If you
figure that out, you get roughly 7 plies of search (5^7 is close to 100000).

Now if your 'selectivity' is good, you search two plies deeper than I do, and
you will find tactical things I don't and win.  If your selectivity is bad, you
will search 2 plies deeper than I do, but overlook something here and there that
causes you to lose instead.

That's how it _can_ play better, because it can go deeper.  But that's how it
_can_ play worse, because it overlooks things that the brute-force program
won't.




>Also, are you saying that one method results in a higher chess rating? If so,
>which do you feel would have a higher chess rating?
>The SSDF (Swedish Rating System) always rates the chess computers with Selective
>Search. From my limited experience playing against both methods, I find that
>Selectve Search does on some occasions make a better move than Brute Force. So
>far I have not found the reverse to be as common. Although I must admit the
>occasions where there is a difference in the move seems to be rather infrequent.


it happens.  Wchess is selective in some way.  And I have sent Dave several
positions in the past where it just 'fell apart' unnecessarily. (We play a lot
on ICC).  He fixes those and gets better.  Nowadays he makes few tactical
mistakes, but sees more with his selective search and is _very_ strong as a
result.




>It appears from the limited response to this controversy that it has not really
>stirred up much of an interest from other club members.
>Mel


it's old news and old discussion.  This has been discussed since the 1950's
in fact.  :)



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.