p. 39 The middle paragraph asserts that fictitious play does not converge to a point,
but rather to a one-dimensional stable manifold. The bottom of the page asserts that
fictitious play converges to a single point, the correlated equilibrium that puts equal
weight of 1/6 on 6 of the pure strategies. The former assertion is correct, the latter
incorrect. If in fact fictitious play converges to a point, it must converge to a
Nash equilibrium; it could not converge to a correlated equilibrium that is not
Nash, such as the one putting weight 1/6 on 6 of the pure strategies. In fact, as
asserted at the top of the play, fictitious play converges to a one-dimensional stable
manifold. This manifold is contained in the set of correlated equilibria, and contains no
Nash equilibria. Play cycles near this manifold, albeit at a slower rate as time passes.
