RANDOMIZATION AND ADAPTIVE LEARNING IN A SIMPLIFIED POKER GAME

Thirty pairs of subjects participated in three different two-person ze rosum Poker games in extensive form with imperfect and asymmetric info rmation. The results provide no support for the (unique) mixed-strateg y equilibrium solution for risk-neutral players on either the individu al or the aggregate level. Compared to this solution, the informed pla yers do not bluff as often as they should, and the uninformed players call too often. Comparison of the present study with previous studies of diametrically opposed conflicts shows consistent differences betwee n the extensive form of two-person zerosum games and its strategically equivalent normal form. An adaptive learning model proposed by Roth a nd Erev (1995) succeeds in tracking the general trends in the probabil ities of bluffing and calling over time and in accounting for the effe cts of size of stake, when it assumes that learning takes place only o n stage games when the players are called upon to act. (C) 1997 Academ ic Press.