mclure%Sri-Unix@sri-unix.UUCP (07/15/84)
whether it is
a win, lose, or draw. The author presents additional information for
constructing a program that will play the winning side and win,
although not in the fewest number of moves. Another chapter presents
an Algol routine that classifies such positions according to class and
presents an optimal, fewest-moves algorithm for producing winning K+P
vs. K play. The authors conclusion is that K+P vs. K is considerably
more complicated than most players suspect and that even masters often
misplay positions in this apparently trivial endgame.
The theory of coordinate squares is explained in a chapter by the
series editor M.R.B. Clarke. Other chapters deal with representing
pattern-knowledge in chess endgames, K+R vs. K+N (considerable new
material here for chess endings theory), yet another analysis of
minimax, a chess combination program that uses plans, detecting mate
without tree search, and a technical description of the MASTER chess
program. In this last chapter there is an interesting treatment of the
K+B+N vs. K endgame in which the author present a 7-ply lookahead
tree-search (not too hard for this endgame) that solves this endgame.
This is the first example I have seen where a program can regularly win
as the stronger side in the K+B+N vs. K endgame. Even Belle in 1981
could not solve this endgame.
All in all, I highly recommend this book. If the other books in
this series are as good as this one, the series is superior faire.
Recently a 4th conference was held at which material was gathered for
volume 4 of this series. We should see it coming out sometime this
year I hope. Besides the ICCA Journal, this is the only ongoing series
on computer chess theory I have seen.
Stuart