Using the Signature Method to Classify Commodities and Select Commodity Options Strategies

Can we distinguish between various commodities markets purely based on the time series of their prices? In this talk; we use the signature method (related to rough path theory) to construct features for a random forest that distinguishes between storable and non-storable commodities with a high degree of accuracy. It turns out that; in the same way that machines can identify dogs in digital images; price formation in a given market has certain tell tale signs. This is true even if we normalize according to volatility. These insights allow us to explain why certain standard options strategies work relatively well in some commodities markets; yet underperform in others.