Online two-way trading: Randomization and advice

Abstract

Abstract We consider the following online two-way trading problem: given some amount of money and a stock (security) of fluctuating prices, we want to perform a bounded number of speculative trades so as to make the most money. We assume a global fluctuation model, where global limits on the minimum and maximum possible prices are given. Previously, optimal algorithms were established in the deterministic case. In this paper we consider two models that improve the competitiveness: randomized algorithms, and algorithms with advice. In most cases we give close to optimal upper and lower bounds on the competitive ratios.