Triggering long-short trades on indexes

This paper analyzes the predictivity and return performance of the Barmish-Iwarere feedback trading algo- rithm described in (1). In the first part of the paper, we study the trade triggering algorithm using either an Ito process model, or real data from indexes and ETFs. It is shown through hypothesis testing that the trigger provides mixed results in predicting the sign of the single trade, for both the Ito process and real indexes. However, we show empirically that the trigger is sufficiently good in identifying a trend, while it fails in detecting side movements. In the second part of the paper, we analyze the effect of controller parameters under various market circumstances. The efficiency of a pre-optimization on historical data appears controversial. Some modifications are experimented, with the objective of improving the returns. In particular, the trigger is modified to detect anomalous falls during a rising trend using the estimated volatility. The resulting system is then tested with other indexes, commodities and interest rates. Index Terms—Trading system; trigger; feedback controller; long-short trades. I. INTRODUCTION A mathematical model which is frequently used to approx- imate the behavior of real markets is the Ito process (2), that is a Brownian motion with drift. A Brownian motion (also known as Wiener process) has three properties: - is a Markov process: the probability distribution for all future values of the process depends only on its current value; - has independent increments: the probability distribution for the change in the process over any time interval is independent of any other (non overlapping) time interval; - changes over any finite interval of time are normally distributed. An Ito process is described by the equation: