Author: Christophe Theron
Date: 12:33:57 04/20/98
Go up one level in this thread
On April 20, 1998 at 09:56:15, Amir Ban wrote:
>I think your question is more likely to be asked by a non-programmer
>than a programmer. The Junior depth value measures half-plies, so if you
>insist, it's ply 6. I could have called it ply 12 or ply 9 with equal
>justification. Since nobody does brute-force anymore, and everybody does
>both pruning and extensions, what the depth indicator means is a bit
>vague. Very roughly, I would propose this equality:
>
> Genius depth 6 = Rebel depth 8 = Fritz5 depth 10 = Junior depth 12
>
>which once said should be forgotten, since it's much more complicated
>than that.
No No No. I don't want to forget that, because I find this point very
interesting. We can talk about it a little bit.
As many people, I have always been confused by the depths and lines that
Genius shows. Some time ago I gave up trying to understand.
In the case of Fritz, it seems quite clear. The depth is the number of
plies where the program is allowed to look for any legal move. This
number of plies is followed by an unlimited (?) number of plies where
the program is only allowed to look at checking moves, check evasions,
captures and promotions (we call this quiescence search).
That's why, even at depth 6, Fritz can show lines with more than 6
moves. But you will notice that the last plies are always captures,
checks, etc.
The depth can be increased by "extensions". I don't know exactly which
are the one Fritz uses. Those extensions create some extra confusion
too.
By playing a lot of games against some of the programs you are talking
about, I came to this conclusion:
Genius depth 6 = Fritz depth 8 = Rebel depth 7
I cannot say anything about Junior, as I don't have it.
Genius depth 6 means, I think, that the program is considering the first
6 plies in a brute force manner (or nearly) and is considering
selectively 4 plies more, followed by a QSearch. So I understand why you
write Genius depth 6 = Fritz depth 10. But I would say that the 4
selective plies of Genius are not very efficient, and are nearly equal,
in quality, to a 2 plies search. That's why I say Genius depth 6 = Fritz
depth 8.
In the case of Rebel, it seems that it is able to understand that some
lines are pretty undecided at the point where they end, and that the
program is able to continue its work deeper in that case. While it is
true that it is able to do it for a large number of plies in some tricky
cases, I would say that it is equivalent, in quality, to a 1 ply search.
All of this takes also into account the different extensions schemes of
each program, because it is based on games I have played.
This topic is interesting and is not reserved to chess programmers. Any
user of these programs could set up a match at fixed depths, and try to
find at which depths the programs play roughly at the same strengths.
They could start with the equivalences given by Amir, or with the
equivalences I'm giving here.
I would really be interested to know the result of such matches. I think
it wouldn't be very long to find out the right equivalences, because you
don't need to set high depths. Fritz depth 8 results in a blitz game.
Any volunteer to help the understanding of this topic? Thorsten?
Fernando? Enrique? Torsten? ... ... ...
Christophe
This page took 0.01 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.