Selecting and Backtesting Cryptocurrency Pairs for Mean Reversion
Summary
This document explains a statistical arbitrage approach that trades two correlated cryptocurrencies when their price ratio moves away from a reference level. It describes taking opposite positions in the two assets and closing or adjusting them as the ratio returns toward its average. The example workflow gathers perpetual contract candles, calculates correlations to select pairs, and uses a simple simulated exchange to track positions, fees, exposure, and account value.
The document reports backtests on four currency groups and says results looked favorable, but provides limited detail about their magnitude or robustness. It acknowledges that selecting pairs with correlation calculated over the full sample introduces future-data bias, and describes splitting data into selection and trading periods as a check. It also flags changing relationships, extreme deviations, low liquidity, and transaction costs as risks, and suggests recalculating correlations, setting exits, and diversifying. Correlation alone does not ensure a stable spread or mean reversion, and the historical examples do not establish live performance.
Key ideas
- Pair trading takes opposing positions in two assets when their relative price departs from a reference ratio.
- The example uses historical correlation to select candidate cryptocurrency pairs and simulates trading costs and exposure.
- Using the full sample to select pairs leaks future information into a backtest.
- Relationships can change, and extreme moves or poor liquidity can cause losses and execution problems.
- Fees from frequent rebalancing can erode any gains from spread convergence.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.