Overview
The Statistical Arbitrage Pairs Trading strategy exploits temporary price divergences between historically correlated cryptocurrency pairs. Using cointegration analysis and machine learning, it identifies when pairs deviate from their equilibrium relationship and trades the expected reversion.
Core Methodology
Pair Selection
Pairs are selected using a multi-factor scoring system:
- Cointegration Score — Engle-Granger and Johansen tests confirm long-run equilibrium relationships
- Correlation Stability — Rolling 90-day correlation must remain above 0.75
- Liquidity Filter — Both assets must have minimum $10M daily volume
- Sector Alignment — Pairs from the same sector (L1s, DeFi, Infrastructure) show stronger mean reversion
Signal Generation
Spread = log(Asset_A) - beta * log(Asset_B)
Z-Score = (Spread - Moving_Avg) / Moving_StdDev
- Entry: Z-Score crosses +/- 2.0 standard deviations
- Exit: Z-Score reverts to +/- 0.5 standard deviations
- Stop Loss: Z-Score reaches +/- 3.5 standard deviations (breakdown)
ML Enhancement
A gradient-boosted model refines entry timing by predicting the probability of mean reversion within 48 hours, using features like:
- Order book imbalance ratios
- Funding rate divergence
- Volume profile asymmetry
- On-chain flow differentials
Active Pairs
| Pair | Sector | Avg Holding | Win Rate |
|---|---|---|---|
| ETH/BTC | L1 | 2.1 days | 84% |
| SOL/AVAX | L1 | 1.8 days | 81% |
| LINK/UNI | DeFi | 3.2 days | 79% |
| AAVE/MKR | DeFi | 2.7 days | 83% |
| ARB/OP | L2 | 1.5 days | 86% |
Risk Controls
- Max pairs simultaneously: 8
- Max allocation per pair: 12.5% of capital
- Pair correlation monitoring: Auto-close if 30-day correlation drops below 0.6
- Portfolio beta: Maintained near zero (market neutral)
This strategy is designed to be market-neutral, generating returns regardless of whether the broader crypto market moves up or down.