Unconventional Refinement for VECM-Based Pairs Trading Strategy through Asymptotic Properties

Abstract We propose a refinement framework for model-based trading strategies that filters out trades likely affected by structural changes, identified through violations of asymptotic properties. The key insight is that such trades, deviating from theoretical assumptions, are less likely to yield reliable model-based profits. We focus on vector error correction models (VECM) and show that their covariance matrices exhibit Frobenius norm convergence. Building on this result, we design an algorithm that detects and removes trades violating this convergence pattern. Empirical results on U.S. and Taiwanese stock markets demonstrate improved profitability and stability. Our approach reduces the adverse effects of structural breaks and can be extended to enhance other model-based trading strategies.