Stock Trading Via Feedback Control: An Extremum Seeking Approach

In finance, the goal of technical analysis is to model stock dynamics to make reliable predictions of future prices. In recent years, Prof. B. Barmish and coauthors have proposed a paradigm shift in which stock trading is reformulated as a control design problem. From such a perspective, unpredictable price variations can be seen as external disturbances and do not need to be accurately modeled for the investment policy to be properly selected. Although very powerful, the tuning of the reactive scheme to stock trading is all but straightforward. In this paper, an Extremum Seeking approach is proposed to directly design the control action from data and make the overall strategy adaptive with respect to trend variations of the stock price. The method is extensively backtested on real stock data.