COPElAND & PRINz | 339
Turing’s section on chess was broadly identical to his chess typescript, although with numer-
ous minor differences. There was, however, one major difference: Game 1 was replaced by a dif-
ferent game, Game 2. The two games began in the same way, but diverged after twenty moves.^25
This raises some tantalizing questions. Why did Turing replace Game 1 with Game 2? How
could White’s twenty-first move be different in Game 2, given that all previous moves were the
same? Who indeed was the human player in Game 2? There is no reason to think that Black
was Glennie, who mentioned playing only one game. Unfortunately the chess writer Alex Bell
introduced confusion by reprinting Game 2 in his book and labelling it ‘Game between Turing’s
Machine and Alick Glennie’.^26 But Glennie won at the twenty-ninth move, and while this is true of
Game 1, where Glennie’s twenty-ninth move achieved checkmate, it is not true of Game 2, where
Black’s twenty-ninth move brought neither checkmate nor even check. White could play on.
Eli Dresner of Tel-Aviv University has suggested a plausible solution to the puzzle:^27
Perhaps Turing produced Game 2 by recalculating some aspects of Game 1, and then making
Black’s last nine moves himself. Some of the position-play values have certainly been recalculated:
the values at each of lines 3, 5, 18 and 21 are different. So it’s possible that while Turing was
reflecting on White’s performance during the Glennie game, he found that he had miscalculated
White’s 21st move, and therefore he reworked the last part of the game. It wouldn’t have taken
him more than about an hour, given that the whole game with Glennie lasted 2–3 hours. Maybe
Turing was even quite pleased to discover that his program lived to fight on after the 29th move!
With Baby growing by leaps and bounds in the Manchester lab, Turing soon had enough
computing power on hand to make it feasible to try out a heuristic chess program. Michie did
indeed recall that Turing made a start on coding Turochamp for the computer. But ‘he never
completed it’, Michie said.^28 Eventually—in the next millennium—Turochamp Mark 2 was
finally implemented by ChessBase. But before we pick up the ChessBase story and move to the
present time, we will describe another of Manchester’s historic chess programs. Its programmer,
a chocolate-loving German physicist, beat Turing to the first working chess implementation.
The Prinz chess program
During his lecture Kasparov noted that ‘a Turing disciple, Dietrich Prinz’ was the first person
to program a real computer to play chess.^29 Prinz (Fig. 31.2) was a computer geek before the
term was even invented, and Turing himself taught this quiet polite German refugee how to
program the Manchester University Ferranti computer.^30 Prinz went on to write a program-
ming manual for the Ferranti that was a model of clarity, unlike Turing’s own earlier and often
perplexing manual.^31 Prinz also continued the work on computer music (see Chapter 23), and
in 1955 he programmed the Ferranti Mark I to play Mozart’s Musikalisches Würfelspiel (musical
dice-game).^32 Prinz used the computer’s random number generator instead of dice, and by
implementing Mozart’s dice-involving composition rules he made the computer compose and
play folk dances.
How Prinz—well mannered, obsessively tidy, meticulously organized, and always punctual—
got on with the untidy and not-so-well-mannered Turing went unrecorded, but the two cer-
tainly had plenty of intellectual ground in common. Turing would have approved when Prinz
maintained that ‘the standard procedure for making babies’, as he described it, was a way of