Rohini Das

We propose to study financial markets in which both the liquidity-providing market maker and the liquidity-demanding market takers are strategic agents. We consider a sequential decision-making framework in the form of a repeated, incomplete-information game to encapsulate the interaction between a market maker and a sequence of recurring traders (a single trader, or a multiplicity of recurring traders). We want to design learning rules and prove tight upper and lower bounds on the regret when both sides of the market are strategic under a realistic feedback model. We also want to investigate whether the time-averaged play converges to some game-theoretic equilibrium and whether the game dynamics leads to an emergence of algorithmic collusion. We also aim to study other theoretical questions in the intersection of learning theory and online learning.