Author: Christophe Theron
Date: 09:43:27 04/27/98
Go up one level in this thread
On April 27, 1998 at 02:00:07, Komputer Korner wrote:
>If you mean that Fritz 5 and Junior 4.6 are root processors, has anybody
>developed a definitive test set of positions that show this? Since the
>amount of knowledge applied to the tips vs the root evaluations is all
>relative, it is difficult to classify root processors as completely
>separate from leaf processors. Nimzo is the classical root processor,
>with Fritz not far behind and I suspect that Junior is also fairly
>close to the root processor end of the scale.
>--
>Komputer Korner
I'm not sure we could design a set of positions to prove that a given
program is doing intensive root processing. After all, we don't know
what is this processing, and it could be different from one program to
another.
However, here is one of the situations where those programs show their
nature: an intensive root processing program has problems to evaluate
correctly the transitions between the game phases.
It is especially the case in middle game positions where a (long) swap
sequence leads quickly to an endgame position. Sometimes you can see it
after the queens exchange. Before the exchange(s), the program considers
the whole swapping sequence and give it a score of, say, A. When the
exchanges have been done, the program suddenly discovers that he is in
the endgame, with a score of B, with B being far less than A.
So you could try to find 5 or 10 middle game positions with a forced
exchange leading to an endgame and try them on Fritz, Nimzo and Genius.
I'm sure that Genius has problems in these positions. I suppose Fritz
and Nimzo would too. I cannot be sure that this kind of test would
always prove intensive root processing, because a well designed root
processor could maybe avoid this phase transition problem. But it is
very difficult.
Christophe
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.