Experimental Analysis of an Online Trading Algorithm

Trading decisions in financial markets can be supported by the use of online algorithms. We evaluate the empirical performance of a threat-based online algorithm and compare it to a reservation price algorithm, an average price algorithm and to buy-and-hold. The algorithms are analyzed from a worst case and an empirical case point of view. The effectiveness of the algorithms is analyzed with historical DAX-30 prices for the years 1998 to 2007. The performance of the threat-based algorithm found in the simulation runs dominates all other investigated algorithms. We also compare its performance to results from worst case analysis and conduct a t-test.