Abstract

In recent years, there has been an explosion of research in metaheuristics, which provides efficient solutions that are close to optimal with lower computing times. Applying metaheuristics to finance is reasonable given that many financial decisions must be made within very short time frames, minutes or even seconds such as in the case of High Frequency Trading. In this paper, an algorithm based on Ant Colony Optimization metaheuristics is proposed to dynamically optimize the decision thresholds provided by the Pairs Trading investment strategy.The proposed algorithm is called the Ant Colony Optimization of Pairs Trading (ACO-PT) and is optimized by moving training-trading windows.The model is applied to Forex data at a high frequency, consisting of 38 Foreign Exchanges with a frequency of 15 min from September 22, 2017 until July 6, 2018. It is shown that ACO-PT can be used in deep markets efficiently and is capable of obtaining daily returns of 0.1204% and a Sharpe ratio of 0.6520, which translates into an improvement over the base case for fixed thresholds of 13.21%. We conclude statistically that the variation of the algorithm that showed the best performance was also the simplest variation and, therefore, the fastest.