Around 1948 Alan Turing, with his colleague David Champernowne, devised a chess-playing procedure they called Turochamp. There was no machine capable of running it at the time, so Turing executed it by hand, following his own rules step by step on paper to choose each move, in what is often described as the first program ever written to play a complete game of chess. He set out the method in a chapter usually titled “Digital Computers Applied to Games,” published in 1953 in B. V. Bowden’s collection “Faster Than Thought” (Pitman, London). The original typescript with his manuscript corrections survives in the Turing Digital Archive at King’s College, Cambridge.
Turochamp was not a brute-force search of all possibilities. It assigned numerical values to material and position, looked a small number of moves ahead, and chose the move that maximized its evaluation, with rules to keep searching down forcing lines such as captures and checks. These are the core ingredients, position evaluation and limited lookahead, that game-playing programs still use today.
The work matters to the prehistory of artificial intelligence because it took a task long regarded as a hallmark of human intelligence and reduced it to an explicit, mechanical procedure. Turing was demonstrating that reasoning about a game could be specified precisely enough to be carried out without insight or intuition, only rules. The primary source used here is the archival record of his typescript.
For a general reader, Turochamp is an early, concrete answer to a question that still matters: how much of what looks like clever human judgment can be captured as a set of rules a machine could follow.