Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Will comp-comp become more positional as computers become faster?

Author: Mike S.

Date: 10:14:15 05/31/03

Go up one level in this thread


On May 31, 2003 at 12:56:04, Reinhard Scharnagl wrote:

>On May 31, 2003 at 12:00:06, Mike S. wrote:
>[...]
>> It has also been said (something like): postional play is just another term
>> for very deep tactics, more or less. This could fit to the play of engines >> when they search very deep.
>
>For my opinion this is not true at all. The problem is still to evaluate quiet
>positions only instead of positionally deescalated ones. You can have only a
>goal within tree evaluation which is visible as a leaf node. As I know evaluated
>leaf nodes today still ignore those dynamic positional aspects. The result is a
>tactically strong play with positional highlights only by luck.

What I wrote was meant from a human viewpoint, not how the engines handle this
"internally". :-)

I was thinking of the *first* move of the pv, which may look like a good
positional move from a human viewpoint, while actually the engine calculates 22
plies ahead and sees that this move will in each case provide a material
advantage or even a mate, then.

In such cases it wouldn't even be neccessary to evaluate quiet positions at the
end of such a search *positionally*, but still the resulting move sequence the
engine finds, may look like perfect positional play from a human viewpoint...

If a positional advantage cannot be proven by tactical means (sooner or later),
than it isn't and advantage at all. Therefore, very deep calculation - which the
inital poster was talking about (= very long calculation time) - can substitute
clever positional evaluations (and can also substitute knowledge).

You could see positional knowledge as an approximation of very deep calculation.
I'm not 100% sure now, but IIRC Tarrasch or Lasker already said that, somewhere.

If an engine could calculate 1.000 plies deep, it would'n need positional
evaluations at all. Because, I guess there's hardly a 500 moves game which isn't
won yet, or drawn by the 50 move rule (to illustrate my point with an extreme
example).

Therefore, it's obvious for me why computer games from very long thinking time
look much more "positional" although the engines use the same evaluations like
in a blitz game...

Regards,
M.Scheidl



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.