Quant Commodities Task by Alpha Alternatives

  1. Statistical Arbitrage:
    • Identified a solid mean-reverting relationship between NG1 and NG2 prices.
    • Generated a Sharpe Ratio of 1.00, indicating moderate risk-adjusted returns.
    • Maximum drawdown was -10.61%, with a recovery duration of 107 days.
    • Momentum strategies were mixed: Time Series Momentum delivered a Sharpe Ratio of 1.35, while Cross-Sectional Momentum performed poorly, with a negative Sharpe.
  2. Machine Learning Approach:
    • Leveraged feature engineering combined with a random forest regressor model.
    • Achieved a Sharpe Ratio of 9.05, significantly outperforming the statistical approach.
    • Delivered higher cumulative returns, showing greater potential for profitability.

Conclusion: While Statistical Arbitrage identified a strong price relationship with reasonable risk-adjusted returns, the Machine Learning strategy far exceeded expectations in terms of performance. However, to build a more robust and risk-managed trading strategy, a combination of the Machine Learning model with a risk-adjusted variant of the Statistical Arbitrage approach is recommended.

Submission Report

Github codebase