Design of Pipeline Framework for Pair Trading Algorithm

Abstract

The pair trading strategy is widely considered as an efficient way to trade two stocks and find the correct instance to purchase and sell the shares. The strategy focuses on two stocks that are correlated such that a change in the price of one affects the other. Until now, there have been no automated processes or algorithms to implement the pair trading strategy. With the recent growth of trading apps and websites like Ameritrade and Webull, the market of trading applications has seen some healthy competition that pushes the development of new ideas. The objective of this paper is to bring pair trading to the application era with the use of Machine Learning techniques and algorithms. The implementation is based on a pipelined structure that feeds the readings from the current calculation to the next one for better accuracy and efficiency. Since the algorithm aims to crunch a large amount of data, it makes sense to test it over a cloud computing platform that trades bandwidth for computing power, like Quantopian, that also has support for financial datasets, which is a plus. Thus, this paper gives an implementation of the pair trading strategy by creating a pipelined framework over python and tested on Quantopian systems, thus achieving a high efficiency and accuracy score.