**Authors:** Masood Tadi, Jiří Witzany
**Published:** _Financial Innovation_, 2025
**Link:** [DOI](https://doi.org/10.1186/s40854-024-00702-7)
---
### Summary
This study presents a #Copula based trading strategy specifically designed for #Cointegration of cryptocurrency pairs. It introduces a reference asset-based copula method to improve statistical arbitrage by analyzing mispricing indexes derived from copula-based dependencies.
---
### Key Points
- **Pairs Trading Strategy:**
- Capitalizes on temporary price divergences between two **cointegrated cryptocurrency pairs**.
- Uses **linear and nonlinear cointegration tests** to select pairs with strong co-movement.
- **Copula Method:**
- Traditional methods like distance and cointegration approaches only capture linear relationships.
- Copulas can model **nonlinear dependencies** and **tail behavior** of asset prices.
- The **reference asset-based method** uses **BTCUSDT** as a stable baseline to form spread processes for each altcoin.
- **Trading Signals:**
- **Mispricing Index:** A calculated value that indicates divergence from expected price relationships.
- **Conditional Copula Probabilities:** Generate trading signals when price relations deviate from the modeled copula structure.
- Uses **Gumbel, Student-t, and Tawn-type copulas** for capturing complex dependencies.
---
### How It Works
1. **Formation Period:**
- Identify cointegrated pairs using #Engle-Granger #EG and #Kapetanios-Shin-Snell #KSS tests.
- Construct #spread processes with BTCUSDT as the reference asset.
2. **Trading Period:**
- Generate long/short positions based on the **copula conditional probabilities**.
- Monitor divergence in real-time and close positions when thresholds are met.
3. **Backtesting:**
- Evaluated on **Binance USDT-margined futures contracts** across **5-min** and **hourly** intervals.
- Showed **205.9% net return** on 5-min data at optimal parameters.
---
### Strengths
- Captures **nonlinear dependencies** that traditional methods miss.
- More resilient to **market volatility**, particularly in #high-frequency trading.
- Outperforms **distance-based** and **return-based copula methods** in terms of profitability.
---
### Limitations
- Relies heavily on BTCUSDT as a reference; shifts in Bitcoin dominance could affect accuracy.
- Higher transaction costs in **5-min data**, despite better returns.
- Requires stable exchange APIs and low slippage for execution.