344 | 31 COmPUTER CHESS—THE fIRST mOmENTS
The Turing–Shannon playoff
Across the Atlantic Claude Shannon, a brilliant mathematician and engineer, also took an inter-
est in computational chess in the post-war years. Like Turing, Shannon was a leading pioneer
of the information age. And playful. He could be spotted riding his unicycle along the cor-
ridors of Bell Labs—sometimes juggling at the same time, it’s said.^54 In early 1943 Turing spent
two months working in the Bell Labs building in Manhattan where Shannon also worked (see
Chapter 18). Possibly he told Shannon about the computational chess ideas that he had previ-
ously discussed with Good and Michie, though no records exist of Turing’s conversations with
Shannon, so we shall never know for sure.
In October 1948, a few months after Turochamp Mark 1 was unleashed on Champernowne’s
wife, Shannon wrote a paper on computer chess.^55 The following year he gave a lecture based on
this paper to a convention of radio engineers in New York, and the paper was finally published
in 1950, titled simply ‘Programming a computer for playing chess’.^56 It sat rather oddly in a
magazine otherwise devoted to hard science, flanked on one side by a paper about the propa-
gation of electromagnetic disturbances along a buried cable, and on the other by a discussion
of the kinetics of phase-transitions in superconductors. Shannon himself took a hard scientific
line about the concept of thinking:^57 after pointing out that ‘chess is generally considered to
require “thinking” for skilful play’, he rammed home the point that the existence of a chess
program with the ability to beat skilful human players
will force us either to admit the possibility of mechanized thinking or to further restrict our
concept of thinking.
It used to be thought that Shannon’s paper and lecture were the earliest work on computa-
tional chess, but thanks to Michie and Good we now know that Turing was pioneering the field
much earlier, during 1941–42. Good said:^58
Basically we came up with the Shannon concept—I thought it was just obvious.
Shannon’s 1950 article contained sufficient detail for Feist to write a corresponding program,
the ‘Shannon Engine’. In an interesting twist, Feist’s Shannon Engine and Turing Engine went
head to head in a ten-game exhibition match, organized by Ingo Althöfer of Jena University.^59
Fittingly, this battle of the Titans ended in a draw. In fact, Turing and Shannon won a single
game each and the remaining eight games were drawn, a draw being declared if the same posi-
tion was ever repeated three times, though Althöfer notes:
In the majority of the drawn games, Turing’s engine had the upper hand.
The match turned into a cliff-hanger. The first excitement came when Turing won the fifth
game at the sixty-third move, after Shannon had blundered in the mid-game: ‘Turing took
the present and never gave back the –6.0 (and better) lead in evaluation’, Althöfer said. When
the end came it was ‘a mate in 4 against the naked White king’. Then draw after draw followed.
Turing seemed about to carry off the match when Shannon suddenly rallied in the middle of
the very last game. At the eighteenth move, Turing had fallen victim to a so-called ‘horizon
effect’, where a move is selected that would have been rejected if only the program had seen
the consequences, which lay a little way beyond its look-ahead horizon. ‘Shannon took the
opportunity with both hands and left White no chance at all’, commented Althöfer. Shannon
won at the thirty-ninth move.