p.7 In order to be able to play chess well, a computer program need only
know the rules of the game and how to evaluate any position which may arise. The rules of the game are not difficult to program,
and are implicit in the legal move generator. What poses the real difficulty is the evaluation of chess positions.
p.8,10 After material, the next most important feature in chess is mobility. Mobility is the total number
of moves that can be made by one's pieces, and this simple count provides an excellent measure of one's freedom of movement...
experiments on the significance of mobility in chess have been carried out as long ago as 1949, when Eliot Slater published
some statistics based on 78 randomly chosen master games... what is interesting is the almost consistent manner in which the
winner's advantage in mobility increases as the game progresses.
p.59 It is clearly useful to devise ways of pruning the game tree. A human chess master can do this quite
simply - he knows that certain moves "just cannot be right", so he ignores them. This reduced the size of his game tree to
something between 50 and 100 positions. For a computer program, anything which helps to reduce the number of moves
in the game tree is a boon, provided that it does not result in the program overlooking something vital... If the
program's evaluation function were sufficiently accurate, it would be possible to reject moves simply on the basis that they
lead to a position which is worse than the position currently under consideration.
p.75 One of the most controversial aspects of computer chess is
the question of whether or not a computer program can eventually be stronger than the (human) World Chess Champion... No-one
knows the answer to this question. My own opinion is that it will be between the years 1995 and 2000 when a computer
program can play at the level of a strong Grandmaster under tournament conditions.
p.76-77 An increase in the number of nodes examined per move is usually
reflected in an increase in the program's tactical ability. A program might well improve tactically by 100 points, but as
a player becomes stronger, tactics play a smaller part in determining the results of his games. At the very top level,
in the World Championship matches (for humans), only a very small proportion of the games are decided by tactics. Hence,
improving a program's tactics will not result in an overall improvements in its play to the same extent.
p.82 The syndrome known as the Horizon Effect manifests itself in positions in which a program cannot see
some crucial move because that move is beyond the horizon of its look-ahead analysis. As a result of this short sightedness,
the program makes a serious error. [JLJ - therefore, machines must check to verify that the position is resilient before making
a search cut-off]
