Kelly-based stock trading via feedback control

In stock trading, Kelly betting has recently gained much popularity as a methodologically sound approach to the computation of optimal investment fractions. In this work, by extending Kelly’s original idea to time series, we develop a novel closed-loop trading investment strategy. The proposed algorithm first estimates a probabilistic model of the price dynamics from the available data, then employs the identified model to compute the investment fraction, resorting to Kelly betting theory. Experimental validation of the algorithm on real-world market quotations shows promising results against state of the art solutions.