Agent-based markets: equilibrium strategies and robustness.

Agent-based modelling (ABM) is broadly adopted to empirically study the market microstructure. Researchers set up market mechanisms and behaviour rules for participating traders, modelled as agents, and observe the simulation results. However, these results can qualitatively change if trader incentives are ignored - an equilibrium analysis is key to ABM. Empirical game-theoretic analysis (EGTA) is widely adopted to compute the equilibria of these agent-based markets. In this paper, we investigate the equilibrium strategy profiles, including their induced market performance, and their robustness to different simulation parameters. For two mainstream trading mechanisms, continuous double auctions and call markets, we find that EGTA is needed for solving the game as pure strategies are not a good approximation of the equilibrium. Moreover, EGTA gives generally sound and robust solutions regarding different market and model setups, with the notable exception of agents' risk attitude. We also consider heterogeneous EGTA, a more realistic generalisation of EGTA whereby traders can modify their strategies during the simulation, and show that fixed strategies lead to sufficiently good analyses, especially taking the computation cost into the consideration.