Experimental Investigation of a Cyclic Duopoly Game

The notion of a cyclic game has been introduced by Selten and Wooders (2001). They illustrate the concept by the analysis of a cyclic duopoly game. The experiments reported concern this game. The game was played by eleven matching groups of six players each. The observed choice frequencies were compared with the predictions of Nash equilibrium, impulse balance equilibrium (Selten, Abbink and Cox (2005), Selten and Chmura (2007)) and two-sample equilbrium (Osborne and Rubinstein(1998)). Pairwise comparisons by the Wilcoxon Signed-rank test show that impulse balance equilibrium as well as two-sample equilibrium have a significantly better predictive success than Nash equilibrium. The difference between impulse balance equilibrium and two-sample equilibrium is not significant. In each matching group three players acted only in uneven periods and the other three only in even periods. This game has two pure strategy equilibria in which both types of players behave differently. The data exhibit a weak but significant tendency in the direction of coordination at a pure strategy equilibrium.