Author: Amir Ban
Date: 01:55:26 01/27/98
Go up one level in this thread
On January 27, 1998 at 02:47:39, Bruce Moreland wrote:
>
>On January 26, 1998 at 18:41:29, mike cooter wrote:
>
>>[White "TJunior46"]
>>[Black "TFritz5"]
>
>>{W=14.0 ply, B=9.1ply}
>
>I assume these numbers are average depth for the whole game.
>
>Someone could look at this and go "wow" in the case of Junior, but I
>have also heard that it is hard to tell exactly how deep Junior really
>searches, since it is keeping track of depth in 1/2-ply units or
>something like that.
>
>Amir, what is going on here? How should people interpret this statistic
>that is being output by your program?
>
>Is Junior really searching five plies deeper than Fritz?
>
>bruce
This statistic is attached to the pgn by the Fritz5 program.
I've answered this in conversations and postings many times, but I'm
surprised to hear this question coming from you. It's still as Shay told
you a long time ago, and as we again discussed in Paris.
I don't have a problem in answering the question technically, and I do,
but what I really want to answer on such occasions is that it doesn't
matter. In my program, it's fair to say that I've completely lost sight
of brute-force. It's very rare to see a PV with only brute-force depth,
or a situation where an unextended line matters. It's much more common
to see a PV as long as the advertised depth or longer. If you want to be
scientific about it, then null-move programs need to subtract at least 2
from their advertised depth.
Regarding the published massacre: I'm seeing such results consistently
on my P5/100 notebook, but much better ones on P5/133 and P6/200. I have
no idea why results should vary so much by platform. In any case Fritz5
is extremely hard to beat in a long match.
Amir
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.