Parameters Optimization of Pair Trading Algorithm

TL;DR

The paper tackles profitable pair trading by identifying cointegrated pairs and trading on the mean-reverting spread $Z_t$ between asset pairs. It proposes a parameter-optimization framework with Point-In-Time validation, using temporal segmentation and Bayesian optimization (Optuna) to optimize entry and exit thresholds $θ_{in}$ and $θ_{out}$. Results on SP500 data show around 872 pairs with correlation above 0.8; baseline cumulative returns of about $5.2\%$ over a 3-month window, with optimization refining thresholds to about $θ_{in}^*≈1.42$ and $θ_{out}^*≈0.37$, albeit with greater tail risk. The work highlights that larger asset pools can improve opportunities and that careful PIT-guided optimization can yield robust profitability, while also noting computational limits and potential future enhancements via machine learning.

Abstract

Pair trading is a market-neutral quantitative trading strategy that exploits price anomalies between two correlated assets. By taking simultaneous long and short positions, it generates profits based on relative price movements, independent of overall market trends. This study explores the mathematical foundations of pair trading, focusing on identifying cointegrated pairs, constructing trading signals, and optimizing model parameters to maximize returns. The results highlight the strategy's potential for consistent profitability even in volatile market conditions.

Figures (10)

• Figure 1: Example of two assets with similar behavior.
• Figure 2: Piece of the dataset prices of S&P500
• Figure 3: Temporal division of different periods
• Figure 4: Stock prices for returns with a correlation of $\rho = 0.95$
• Figure 5: Stock pair with significant cointegration test ($\rho=0.93$ and $p$-value $=1.7e-18$)