VIX constant maturity futures trading strategy: A walk-forward machine learning study

Area of Science:

• Quantitative Finance
• Computational Finance
• Machine Learning Applications

Background:

• VIX constant-maturity futures (VIX CMFs) are crucial for hedging and speculation.
• Predicting VIX CMFs returns is challenging due to market volatility.
• Term structure information offers potential predictive insights.

Purpose of the Study:

• To evaluate the predictive power of VIX CMFs term structure for next-day returns.
• To develop and test a novel trading strategy based on machine learning predictions.
• To compare the performance of machine learning models and the proposed strategy against benchmarks.

Main Methods:

• Employed seven advanced machine learning models for numerical predictions.
• Utilized three feature sets, incrementally incorporating VIX CMFs term structure.
• Implemented a walk-forward expanding-window methodology over an 11-year period.
• Proposed and backtested a Constrained-Mean-Variance Portfolio Optimization (C-MVO) strategy.

Main Results:

• Four out of seven machine learning models achieved a prediction information ratio > 0.02 (average 0.037).
• VIX CMFs term structure features demonstrated significant predictive power.
• The C-MVO strategy yielded an average information ratio of 0.623, outperforming the benchmark long-short strategy (0.404).

Conclusions:

• Machine learning models effectively predict VIX CMFs next-day returns.
• The VIX CMFs term structure contains valuable predictive information.
• The proposed C-MVO strategy, powered by machine learning, offers superior risk-adjusted returns.