[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Bug-gnubg] Predefined move filters

From: Joseph Heled
Subject: Re: [Bug-gnubg] Predefined move filters
Date: Tue, 31 Dec 2002 11:03:23 +1300
User-agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.2) Gecko/20021202

Gary Wong wrote:
On Mon, Dec 30, 2002 at 07:11:30PM -0200, Albert Silver wrote:

I'll do some testing of this, but it's not always easy to see the
difference. One detail: It's true that the 1-ply is the worst (IMHO) ply
depth, however, in my experience, the 3-ply is definitely stronger than
the 2-ply. This has been confirmed by numerous rollouts, as well as
errors pointed out by Snowie.

I agree.  In fact, I think 1-ply comes out worst only when the judge
is an even-ply evaluation; head-to-head, 1-ply beats 0-ply by a tiny

I disagree. 0ply beats or equals 1ply when the judge is 0ply rollouts, as my benchmark shows for both race and crashed positions.

And trying to specifically improve the 1ply results in degrading the 0ply. Again that might reflect the specific methods I use - as yet I have no idea what can be changed to make 1ply better without degrading 0ply.


The ply oscillation problem is particularly annoying.  It's
reminiscent of all those court cases where decisions get appealed from
court to court, being reversed at each step along the way, and the
final outcome is really a matter of which side was lucky enough to be
"in phase" when the ultimate court is reached, rather than whatever
merits one might find to consider.  (Of course, in the U.S.A., they
pick presidents that way.)


reply via email to

[Prev in Thread] Current Thread [Next in Thread]