332 | 31 COmPUTER CHESS—THE fIRST mOmENTS
play a game of chess against a beginner and stand a fair chance of winning or at least reaching a
winning position.
Turochamp was (so far as we know) the first chess program ever to beat a human player—
Champernowne’s wife. There is no record of either Turochamp, or indeed the later chess pro-
gram that Turing wrote, ever beating Turing himself, but he likened the claim that no program
can outplay its own programmer to the claim ‘No animal can swallow an animal heavier than
itself ’.^14 ‘Both’, he said, ‘are, so far as I know, untrue’.
A few weeks after these historic experiments with Turochamp, Turing took up his new job at
the Manchester Computing Machine Laboratory, putting him almost within reach of coding a
chess program for a real computer. Things didn’t quite work out, however. But before relating
that part of the story we will describe a distinction that is essential for understanding the full
significance of Turing’s achievements in chess programming.
Brute force versus heuristics
Chess programmers talk about ‘chess heuristics’: the term ‘heuristic’ is explained in Chapter 25.
Chess programs using heuristics are contrasted with ‘brute-force programs’. A brute-force chess
program selects its moves by exploring the consequences of each move available to it, even right
through to the end of the game. If programmed in this way, the computer would always choose
the best move.
In practice, though, a chess program that uses nothing but brute force is a hopeless propo-
sition. As Max Newman pointed out in a postwar radio talk, the brute-force method would
only work ‘if you didn’t mind its taking thousands of millions of years’.^15 This is because the
number of moves that the computer would have to examine is astronomically large. In his war-
time discussions with Good and Michie, Turing had hit on the more practical idea of playing
chess by using heuristics. Instead of examining every available move exhaustively, the machine
would use rules of thumb—supplied by the programmer—to select moves for further analysis.
Turochamp seems to have been the first heuristic chess program.
These rules of thumb are like shortcuts, enabling the computer to play quickly—but the price
of speed is that shortcuts don’t always work. When a shortcut happens to lead in the wrong
direction, the chess program may make a mistake. No doubt Turing was alluding to this feature
of chess heuristics when he said that it is ‘possible to make the machine display intelligence at
the risk of its making occasional serious mistakes’.
Turochamp mark 2
While he was working on his groundbreaking theory of morphogenesis in the Manchester
Computing Machine Laboratory, Turing also found time to develop an improved form of
Turochamp. We will call this ‘Turochamp Mark 2’, and from now on we use the term ‘Turochamp
Mark 1’ to refer to the original 1948 version. Turing described the Mark 2 in a far-sighted type-
script about computer chess (published in The Essential Turing). He explained that the system
of rules set out in this typescript was ‘based on an introspective analysis’ of his own ‘thought
processes when playing’—although with ‘considerable simplifications’. He said modestly: