Chess and Machine Intuition by George Atkinson
Consider now this material from "Chess and Machine Intuition" by
George Atkinson.
How is it that a chess program with a simple evaluation function
(using small amounts of knowledge) is able to play a powerful game of chess? Atkinson gives his opinion. Our search function
can correct, to some degree, for inexact focus by a knowledge-weak evaluation function by eventually stumbling on
the correct sequence in a large search tree.
p.46. "In Turing-Shannon type programs, all chess knowledge beyond
the definition of legal moves is contained in the evaluation function. Even though a large number of terms representing such
positional features as material, mobility, and king safety might be made part of a static evaluation function, the very much
greater number of features recognizable as special cases by expert players cannot even be enumerated, much less included.
In principle, this is a minor defect, for the necessarily crude evaluations are in a sense made more precise by examination
of a sufficiently large tree of potential positions and the use of minimax."
How many moves, in a typical position, does a chess master typically
regard as good?
"The studies of de Groot revealed that the average number of moves
that a master examining a position from a well-played game would regard as good is about one and three-quarters."
How do skilled players see chess pieces?
p.74"One of his [Alfred Binet, famous French psychologist] first
observations was that chess players do not perceive chessmen as having any particular form, but view them rather as symbols
characterized by their individual moves and by their significance at a particular point of the game... The capable player
does not, for example, perceive a Knight as a carved horse's head on a pedestal, but rather as a piece with certain capacities
that serves a particular function in its current position... The skilled player invariably perceives a chess piece in terms
of its significance to the present course of the game."
How do chess masters and amateurs differ in their ability to analyze
newly shown positions?
p.77"During the experiments, de Groot [famous Dutch psychologist]
was impressed by the chess master's rapid grasp of the possibilities in a newly shown position. He observed a vast difference
between master and amateur in the amount of time taken to recognize relevant structure and dynamic pressure. The master seems
to perceive the critical forces almost instantaneously and can immediately suggest specific, appropriate board action."
Advanced chess players seem to look beyond pieces and squares and
see an arena, a battlefield where forces clash, pressures are applied, and events transpire or threaten to transpire due to
long calculated move sequences:
p.79"[Alfred] Cleveland [psychologist] had also remarked on the
perceptual metamorphosis that takes place with gain of experience. As more chess situations take on meaning, pieces are gradually
transformed in the player's mind from static objects to forces that can be exerted. The expert no longer sees discrete
squares and pieces or even abstract counters, just as the competent reader no longer sees individual letters... The board
has become an arena of overlapping zones of significant activity, regions in which certain events are taking place or about
to take place, and in which the impossibility of other events can be exploited. A position is seen in terms of pressures,
of 'fields of force' that require additional force to be applied in a particular region to maintain equilibrium, or to divert
the opponent's influence."
