BUILDING A MEAN-REVERTING PORTFOLIOBY MAXIMIZATION OF SPEED OF MEAN-REVERSIONOF ORNSTEIN-UHLENBECK PROCESS

A method of building a mean-reverting portfolio is proposed, based on estimating the
vector of a stationary linear combination of non-stationary random processes by maximizing the
speed of mean-reverting parameter of the Ornstein  Uhlenbeck process. Compared with
previously known approaches, the laborious procedure for determining the order of vector autoregression is excluded, which simplified and made more reliable the algorithm for constructing a stationary portfolio. Based on the proposed method, a trading strategy for statistical arbitrage was developed, which was tested on data of industry ETF prices for the period
20092019. The efficiency of the strategy and the achieved performance characteristics were
demonstrated: CAGR 5.05%, Sharpe Ratio 1.32 and CAGR 2.93%, Sharpe Ratio 9.15. The
profitability of the strategy is ensured regardless of the state of the market, an indicator of which
the S&P500 index was considered.