Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Fine tuning the engine's strength

Author: KarinsDad

Date: 13:49:43 07/27/00

Go up one level in this thread


On July 27, 2000 at 16:21:51, John Coffey wrote:

[snip]
>
>My goal here was not to emulate human 1200 players but instead being able to
>fine tune the ability of the computer.  Fine tuning implies that the change from
>one level to the next is subtle.  The cheapest, simplest and most consistant
>way to do this is to limit the number of nodes the program looks at.  I also
>think that it is the most desirable approach, because it gives the user a
>finite standard by which he can judge his ability.  (Also given a choice
>between a program that is going to make a bad move quickly and a program
>that is going to think for a long time and make a bad move then I would prefer
>the former.)
>

With this approach, you have to remember how the search engine works with
respect to ply and how it searches exponentially. Also, for a given engine, you
may have to take into account average searched child nodes (there's a term for
this, but I cannot remember it off the top of my head).

For example, in the ply solution might result with a given program on a given
system in:

Ply  Nodes  Time (secs) Rating
 1      35   nil          300
 2     245   nil          900
 3    1505               1200
 4    9065   0.0         1450
 5   54425   0.1         1700
 6  326585   0.6         1900

The rating here is almost (not quite) linear once you get past a certain ply.
Since it is generally recognized that doubling the speed equals doubling the
nodes equals 30 to 60 ELO, it might be possible to use your type of node
reduction equation to come up with fairly decent opponents on the middle to
upper scales. On the lower scales, I think you will have to modify your equation
to correspond to the exponential nature of ply search. In other words, you have
to get to at least ply 4 to correspond to any level of semi-decent play where
you are not hanging pieces right and left and it is in these lower rating levels
that picking the best of x moves type of philosophy becomes more important (e.g.
at 2700, best of x would always be best of 1). Otherwise, you are not coming
close to emulating play at a low rating level, you are just having a random take
and hang piece program (my program had this problem early on when I limited it
to low ply searches and still had some significant bugs in it).

KarinsDad :)



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.