A Cointegration and Clustering-based Approach to Pair Trading of Stocks
from Selected Sectors of the Indian Stock Market
Abstract
This work presents a cointegration-based pair-trading strategy for
identifying stock pairs with substantial cointegration in their prices
across four years (January 1, 2018, to December 31, 2021). After the
Cointegrated pairs are determined, pair-trading portfolios are created,
and portfolio performance is tracked during a one-year test period
(January 1, 2022, to December 31, 2021). Suitable trigger points are
determined utilizing a very powerful spread detection system, allowing
both stocks’ short and long positions to be precisely recognized. The
yearly return of a portfolio is used to assess its performance. First,
twelve sectors of stocks from the National Stock Exchange (NSE) of India
are selected. According to the NSE’s monthly report for the month of
December 31, 2022, the top ten stocks in terms of their free-float
market capitalization from the twelve sectors are selected. Pair trading
portfolios are built using pairs from each sector that demonstrated
cointegration of close prices from January 1, 2018, to December 31,
2021. The portfolios are evaluated based on their return from January 1,
2022, to December 31, 2022. Furthermore, clustering-based techniques
were used to identify stocks that behaved similarly for the period of
four years i.e., January 1, 2018, to December 31, 2021. This was
performed by using three clustering techniques and the best technique,
based on their respective results, was chosen to identify the clusters
to verify the cointegrated pairs. Henceforth, the pairs which were
common in both cointegration, and clustering techniques were regarded as
the most recommended pairs for trading. The work makes three distinct
contributions. First, the paper provides a cointegration-based pair
trading strategy for stock portfolio creation, that can be used to earn
profit by the investors in the stock market. Second, the pair-trading
models are trained and tested on real-world stock market data, with the
results displayed to illustrate the models’ efficacy. Finally, since the
stocks utilized in the pair trading portfolio designs are drawn from
various NSE sectors, the outcomes of the pairings are an excellent
signal of the possible profit that investors could make if they invest
in those sectors using the recommended pair-trading technique.