Deep Value Investing with Hidden Markov Models

Abstract

Historically, algorithmic trading has been the realm of high-frequency, technical analysis-based strategies. This paper aims to see if it is possible to use a trading algorithm for a fundamental analysis-based, “deep value” strategy. The paper’s algorithm first uses credit ratings to identifying distressed companies and then return data on them. From there, Hidden Markov models are used to output buy signals. Then, based on these buy signals, a portfolio is created to maximise Sharpe ratio. Due to the large variety of parameters that the algorithm receives, it was rather inconsistent. Occasionally, if one found the ideal tuning parameters, it could consistently outperform the market. But if one did not manage to input the ideal parameters, the performance could be quite poor. The model could be greatly improved if it was streamlined to be more computationally efficient and optimise certain parameters on its own (so that this would not be subject to the user’s choices).