Abstract
In this paper, we developed an adaptive copula-based pairs trading framework for cryptocurrency markets and examined the performance of a market overlay variant. Pair selection was conducted using cointegration techniques based on the Augmented Dickey-Fuller (ADF) and Kapetanios-Shin-Snell (KSS) tests. Dependence between paired assets was then modeled using copulas, enabling flexible and nonlinear joint behavior. From the estimated copula models, we constructed a Copula Mispricing Index (CMI) trading signal, with closed-form conditional distributions derived under Gaussian, Student-t, Clayton, Gumbel, and Frank copulas. Copula family selection was guided by the Akaike Information Criterion (AIC) and further validated through goodness-of-fit diagnostics based on the Rosenblatt transform, employing Kolmogorov-Smirnov and Cramer-von Mises uniformity tests. The proposed framework was evaluated using an extensive backtesting exercise on 26,257 hourly observations of ten Binance USDT perpetual futures contracts, spanning January 2021 to December 2023. Empirical results indicated that, under baseline transaction costs of 0.08% per round trip, market-neutral copula-based strategies exhibited relatively low risk, with maximum drawdowns remaining below 20%. However, these strategies generated negative net returns after accounting for trading costs. In contrast, the Alpha Overlay variant, which relaxed strict market neutrality by incorporating directional exposure, closely tracked the buy-and-hold benchmark over the sample period. Overall, the findings highlight the robustness and limitations of copula-based pairs trading in highly volatile cryptocurrency markets, underscoring the importance of transaction costs and strategy design in determining net performance.