Computer Chess Club Archives


Search

Terms

Messages

Subject: Generalizability across test sets: tactics and strategy

Author: Joseph Ciarrochi

Date: 02:30:50 12/17/05


My question is: to what extent do engines vary in strength depending on the
opening type 9e.g., highly strategic versus highly tactical). There are alot of
assumptions about some programs being highly strategic and others being
tactical. I would love to see the evidence for these programs being better at
certain openings, and worse at others (and you really need a basement full of
computers to do this test).

I am only about an 1800 player, so what i'm about to say might be off,
but.....It seems to me that tactics and strategy are inexorably linked. Good
strategic programs like rybka seem to utilize strategic knowledge to increase
the chance of a tactical shot.

So you seem to have two key factors here: 1) the likihood of detecting a tactic
(given a certain amount of time) and 2) the liklihood of creating a position
where a tactic appears.

If you test a program in terms of speed to detect a tactic, you are estimating
factor 1 but not factor 2.

Given the inexorable link between tactics and strategy, are there really
programs that are good at closed, strategic positions, but not so good at open
tactical positions, and vice versa? Surely the amount of games in the CEGT would
allow you to test this possibility?


best
Joseph







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.