Explainable AI in Futures

Using daily returns on 34 futures contracts over the 2010-2022 period, we examine the factors driving these returns. We show that all commodities can be grouped by their drivers into intuitive groups based on their factorization into 1) food, 2) metals and oil and 3) precious metals. The three groupings have distinct drivers that determine the commodity futures pricing.We further construct a trading strategy based on the factors computed quarterly and the strategy on the 2010-2022. In an out-of-sample setting, our strategy delivers positive Sharpe ratios on 26 out 34 commodity futures contracts and outperforms commodities futures vanilla buy-and-hold strategy in 22 out of 34 commodity futures contracts. On average, our factor-based strategy delivered 2.15% annualized return premium and 0.095 Sharpe ratio premium over the vanilla futures buy-and-hold strategy out-of-sample. We separately test commodity futures prices on the U.S. market returns and that while contemporaneous futures returns significantly depend on SPY, adding SPY to the optimal factors in prediction of forward looking returns on futures actually decreases the performance of prediction