Score contribution per author:
α: calibrated so average coauthorship-adjusted count equals average raw count
When learning to play a game well, does it help to play against an opponent who makes the same sort of mistakes one tends to make or is it better to play against a procedurally rational algorithm, which never makes mistakes? This paper investigates subject performance in the game of Nim. We find evidence that subject performance improves more when playing against a human opponent than against a procedurally rational algorithm. We also find that subjects learn to recognize certain heuristics that improve their overall performance in more complex games.