Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: 3 interviews about engine protocols with T. Mann, R. Hyatt and M. Blume

Author: Russell Reagan

Date: 21:09:53 08/15/02

Go up one level in this thread


On August 15, 2002 at 23:00:36, Will Singleton wrote:

>Let's assume that I discovered some magic bullet pruning method, and it shrank
>my tree in half.  Sure, it would play better, but the opening book would still
>suck, the crappy eval would still be there, and the code would still be riddled
>with all kinds of bugs.  In short, it would still lose to the commercials.
>
>So, I don't believe one can assert that the fact no amateur program has
>surpassed Crafty by a significant margin is evidence that new techniques aren't
>being used by amateurs.  You may well be right, but that argument isn't
>persuasive.

I think you make a good point here. If a program was superior (even if only
slightly in some areas) in every area of the program, and it only used the
techniques that are public knowledge, that program would be stronger than
another program that was lacking in even one area.

I think it's possible that some of the commercials don't use any special
secrets, but I know for a fact that some do. As I said in another post, Junior
doesn't use null-move, Rebel used to not use null-move (I think Ed might have
added it in Rebel XP, not sure on this one), Cristophe said that null-move isn't
Tiger's main forward pruning method and that he has a number of others that he
uses in combination. So there are three top commercial engines that have some
kind of other secrets.

You are also right that these might not be "big" secrets. If an engine had all
of the well known methods highly refined after years of testing (except for
null-move) and had one small secret added in, it might make a big enough
difference to jump from strong amateur to commercial.

I think their secrets aren't "huge" secrets, like the move from min-max to
alpha-beta, or the discovery of null-move initially, but I think they are still
fairly significant. I might be overestimating it because their engines are also
highly refined in every other area.

I'm glad we discuss things like this, because I get to see that other sides of
the argument are possible :)

Russell



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.