Evolutionary Design of Strategic Behavior in Continuous Double Auctions
speaker: Marco LiCalzi (Università Ca' Foscari Venezia)abstract: We study the emergence of strategic behavior in a continuous double auction with an equal number n of buyers and sellers. We assume that the strategy space of each trader is one-dimensional, but otherwise unrestricted: in particular, mixed strategies are not ruled out. The evolution of strategic behavior is modeled as a learning process driven by a genetic algorithm. Using computer simulations, we provide evidence for the following statements. As the market grows in size, allocative inefficiency tends to zero and equilibrium performance converges to the competitive outcome. The main result is that, as n increases, the equilibrium strategies selected by the learning process morph from (approximately) linear to (approximately) stepwise functions. The nature of the continuous double auction is altered by the number of players in a fundamental way, forcing traders’ strategies to change shape in reaction. Finally, we compare the performance and discuss the stability of evolved strategic behavior against other equilibria proposed in the existing literature.
timetable:
Sat 3 Oct, 15:00 - 15:45, Aula 3
documents:
licalzi
<< Go back