Statistical Arbitrage using High Frequency Pairs Trading - An algorithmic strategy in the US equity market

Abstract

Using three-month minute-by-minute data from S&P 500 constituents we examine and report the return of a high frequency trading strategy by creating an algorithm based on pairs trading. The algorithm is created based on two different approaches and is implemented using a broad set of portfolio configurations to compare the effect on trading profits. We show that high frequency pairs trading generates large positive excess returns before accounting for transaction costs and find that the returns express no significant traditional risk factors. Furthermore, testing of certain portfolios to estimate break-even points in respect of transaction cost and return is conducted. Finally, we implement an approach based on implied volatility to test pairs trading under different market conditions.