Asset Picking Based on a Markov Chain Modeling the Asset Performance
Résumé
In this article, we investigate a new method for selecting assets (stocks/funds) for portfolio management. We first define three states of asset performance (with regard to a benchmark): out-performance, intermediate performance and under-performance. The mathematical model for performance is a Hidden Markov Model (HMM, with the three states), which is well suited to performance modeling. In fact, the reasons why an asset performs well are not necessary assessable since they include both rational features and human biases, such as momentum effects. A different return pdf is estimated for each state through a Gaussian Mixture Model (GMM), allowing the modeling of skewed distributions. Based on these mathematical models of performance and returns, we derive a quantitative criterion for asset picking. This criterion can also be reversed for short selling purposes. Applications on simulated data and historical data show the relevance of both our model and our method in the asset selection process.