Perpetual contract pairing trading strategy, unstable pairing performance problem

At the end of August 2024, on Binance, using the perpetual contract data 2024.1- 2024.8.1, backtesting using the pair trading strategy, I obtained several test pairs with Sharpe ratios >2, or even >3, or 4.

When backtesting at the end of October 2024, it was found that most pairs were no longer profitable.



This shows the following phenomenon

1 Changes in market style have led to the instability of the pair profit relationship.

2 The real-time robot developed using Binance's simulated account token has no opening and closing records on the server. This further illustrates phenomenon 1



Question: What methods can be used to solve the above problems?

1 Increase the frequency of backtesting from once every 2 months to once a week?

2 Use principal component analysis (PCA) to increase the stability of the pair?

3 Others?



Thanks

Hello Rong, 



As you identified, the instability of pair trading strategies is a well-known challenge, often due to changes in market conditions that affect the cointegration between pairs. This is particularly evident during periods of heightened volatility, where previously stable relationships may break down.



To adapt more quickly to these shifts, increasing the frequency of backtesting from once every two months to a weekly schedule could allow for earlier detection of style or trend changes, enabling more responsive adjustments to your strategy.



Are there any other ways to better handle the impact of market shifts?



Yes, you can do this by managing the capital allocation. For this, you may consider implementing volatility targeting. By adjusting position sizes based on the volatility of the spread between pairs, you can manage risk more effectively. During periods of high volatility, reducing position sizes can help limit potential drawdowns, while increasing them in lower volatility periods allows you to capitalise on more stable opportunities.



As you mentioned, applying Principal Component Analysis (PCA) is also a useful approach to enhance pair stability. PCA can help you identify pairs with more robust relationships, reducing sensitivity to market regime shifts. By selecting pairs that contribute less to the principal components with high variance, you can focus on pairs that tend to exhibit more stable cointegration.



To further optimise your strategy, mean-variance optimisation (MVO) can be employed to refine capital allocation. This technique balances expected returns and risk among pairs based on historical backtesting data. While backtesting provides insight into a strategy's past performance, MVO optimises capital distribution across pairs to maximise returns relative to risk, offering a more robust risk management framework.



Please feel free to share any more doubts you have on this. Happy learning!