Testing and Backtesting an Aluminum Producer–Natural Gas Mean-Reversion Strategy

Summary

The article evaluates whether an aluminum producer’s equity and a natural gas ETF could form a mean-reverting pair, based on the role of gas in aluminum production. It tests adjusted price series with a cointegrated Augmented Dickey-Fuller procedure, estimates a static hedge ratio by linear regression, and finds insufficient evidence to reject the null of no cointegration. It nevertheless proceeds to describe a QSTrader implementation as an example of testing a weakly related pair.

The strategy forms a weighted spread, calculates its rolling mean and standard deviation, and uses Bollinger-style z-score thresholds to enter long or short positions and close them as the spread moves back toward its mean. The example uses a 15-bar lookback, entry threshold of 1.5, exit threshold of 0.5, and a hedge ratio of 1.213. These are presented as arbitrary choices. The article explicitly warns that estimating the hedge ratio on the backtest sample introduces lookahead bias, so reported performance would be overstated and a proper evaluation needs separate in-sample and out-of-sample data.

Key ideas

• A plausible fundamental link between two assets does not establish that their prices are cointegrated.
• The CADF test on the example spread did not provide sufficient evidence of cointegration.
• Rolling spread z-scores can define mean-reversion entries and exits using separate thresholds.
• The hedge ratio and signal parameters are fixed for illustration rather than validated through optimization.
• Estimating the hedge ratio on the backtest period creates lookahead bias and exaggerates performance.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.