Learning Behavior in an Experimental Matching Pennies Game

Perceiving Strategic Environments: An Experimental Study of Learning Under Minimal Information

“…18 For the last 10 periods, the frequency at which CO players choose PEC increases to 77%, while 74.5% of RI players, 66.5% of the RI+ players, but only 54% of the LI players choose the PEC. The frequency of PEC for CO players significantly exceeds the frequencies for LI players and RI+ players, 19 while there is no significant difference between the frequency for RI players and CO players. 20 Figure 3(i) shows the development for the 5 period averages of the PEC across the third phase.…”

“…During the last decade, there has been an increasing economic literature on individual learning theories (e.g., Crawford, 1995, Mookherjee & Sopher, 1997, Fudenberg & Levine, 1998, Huck et al, 1999. Those theories apply algorithms to predict individual adaptation processes.…”

Behavioural Game Theory: Thinking, Learning and Teaching

“…EWA and self-tuning EWA generate simulated profits of 879-882, which is only an improvement of 5 per cent over 837, but is 80 per cent of the maximum possible improvement from actual payoffs to clairvoyant payoffs. 45 Partow and Schotter (1993), Mookerjee and Sopher (1994), Cachon and Camerer (1996). 46 Sophistication may also have the potential to explain why players sometimes move in the opposite direction to that predicted by adaptive models (Rapoport, 1999), and why measured beliefs do not match up well with those predicted by adaptive belief learning models (Nyarko and Schotter, 2002).…”

Coordination in Networks Formation: Experimental Evidence on Learning and Salience

“…We find little evidence of convergence to the wheel networks through learning dynamics, while we see some emergence of empty networks in the bi-directional model, which was instead not documented by the previous experiments of Falk and Kosfeld. We also study various learning rules which subjects could have used in the experiments 2 . We in particular compare the Cournot Best Response hypothesis taken by Bala and Goyal in the dynamic analysis of their networking games, with alternative learning rules based on models of Fictious Play (as in Fudenberg andLevine 1998, andCheung andFriedman 1997) and of Reinforcement learning (as in Roth and Erev 1995, and Mookherjee and Sopher 1994. Various experiments have been conducted during the last decade on the same learning models (see Camerer 2003, for a thorough review).…”

“…3 These among others include: experiments on the matching pennies game (Mookherjee and Sopher 1994), the hawk-dove, the stag hunt, the buyer-seller and the battle-of-the-sexes games (Cheung and Friedman 1997), the ultimatum game (Harley 1981, Roth andErev 1995), the beauty contest game (Camerer and Ho 1999), several public goods games (Roth andErev 1995, Chen andTang 1998), and, more generally, games with mixed strategy equilibria (Tang 2001, Camerer andHo 1999), constant-sum games (Mohkerjee and Sopher 1997) and coordination games (e.g., Boylan and El-Gamal 1993, Crawford 1995, Broseta 2001 2 The Model by Bala and Goyal: equilibrium theory Bala and Goyal (2000) propose the following model of non-cooperative networks. Let N = {1, ..., n} be a set of agents, with n ≥ 3.…”