1 Introduction

One of our main flavours of commodity alpha we like to harvest is carry. We make profit from carry by taking a bet that the shape of the commodity futures curve will become more contango or remain uncahnged. Commodity futures curves are divided by obstacles to intertemporal arbitrage. The costlier the storage, the greater is the division and the variability of carry. The segmented commodity futures curve is shaped by four factors:

• Funding and storage costs,
• Expected supply and demand imbalanced,
• Convenience yields and
• Hedging pressure.

Under normal conditions commodity producers take short futures positions in the deferred parts fo the commodity futures curve in order to hedge against price drops. The investor or speculator that offers this insurance is paid a premium and takes a long position in the futures contract. This positive premium comes in the form of the carry premium. On the other hand, commodity consumers take long futures positions in neared dated contracts in order to hedge against unexpected future price surges. The investor or speculator that offers this insurance receives a premium for taking up the risk and takes a short position, in which case contango arises

2 Side of the bet

Throughout we denote the front vs three month spread as F3 and the front vs six month spread as F6. When we refer to positive carry we imply that the spread becomes more contango or less backwardated. For every commodity and tenor, we use the ETF-trick to show the value of a $1 investment at the start of the time-series. If the value of the investment increases it implies the spread became more contango. If it decreases the spread became more backwardated. The plot below gives a bird’s eye view of these return streams. Some of the series are choppy such as those of DF and JO, while others are smooth and almost linear. Most of the carry returns are positive over the long run but there are a couple of exceptions, the most prominent one being QW.

For the majority of commodities above we choose the side of the bet to be short the calendar spread, i.e. we take up a short position in the near and a long position in the far-dated contracts. In this way, we will make a positive return in the event that the spread turns more contango as time progresses. The plot below zooms in on the value of $1 invested in the QW calendar spreads. Notice that both parts of the curve show a clear negative long-term trend. This implies that the QW spreads prefer to be in backwardation.

Below we zoom in even further and consider the value of $1 invested at the start of 2015. Notice that both parts of the curve would have given you a return of between 15 and 20% over this term.

It is this long term behaviour of QW that forces us to choose a default side of long the calendar spread for QW, i.e. long the near and short the far dated contract. In this way we can be profitable if the spread becomes more backwardated.

In the base flavours of carry the default sides of the bets are short all commodities and spreads except for QW in which we take long exposure by default.

3 Size of the bet

This section will take up a more time to complete and I will do so in due course. Roughly speaking it entails the use of meta-labeling to create training samples for our machine learning models to determine the probability of a profitable trade. The positions are then sized with respect to this probability together with the volatility of the underlying calendar spread.