Abstract
In financial markets, stock trading techniques usually require to learn a model of the non-stationary dynamics of the underlying asset in order to make reliable predictions and take effective decisions. In the recently introduced feedback approach to trading, the stock price is instead treated as an exogenous disturbance to a feedback loop scheme and a controller is synthesised with the unique goal of minimising the impact of the return variations on the investment gain. Since such an approach is intrinsically model-free, the tuning of the controller represents a critical task. In this work, we propose a data-driven adaptive control strategy, exploiting the knowledge of the gain/loss as well as the measured returns over a moving window. The proposed approach is extensively back-tested on real-world financial series and the related performance is compared to that of classical feedback schemes and to benchmark investment strategies.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Notes
1 According to this selection principle, the considered benchmark strategies turn out to be rather aggressive, since they imply, for each time instant, the investment capacity saturation for the reference position.