Distributed Mean Reversion Online Portfolio Strategy with Stock Network

Online portfolio selection is a practical problem in financial engineering and quantitative trading. Many empirical studies show that stock performance in the market is likely to follow mean reversion, and strategies based on mean reversion show better return performance than the market average. However, the existing mean reversion strategies are not universal and short selling is not allowed, which is unsuitable for real-time investment. In this paper, we propose a distributed mean reversion online portfolio strategy through a stock correlation sub-network to solve these problems. Theoretical analysis shows that our strategy is universal and the convergence rate is calculated. The empirical results show that our strategy is better than the existing universal strategies in terms of return performance, nor is it sensitive to transaction cost.