Commodity Cross-Sectional Momentum

Thesis (edge)

This is momentum applied inside the commodity complex only. You are not asking whether commodities as a group are going up. You are asking which commodities have been beating the others, and you buy those while shorting the ones that have been lagging.

The economic story is about slow physical adjustment. If copper demand rises, you cannot conjure a new mine in a quarter. It takes years. So a demand shock in copper produces a price rise that persists, because the supply response is slow and everyone can see it coming but nobody can do anything about it quickly. The same is true in reverse: a bumper harvest depresses grain prices, and it takes a full growing season for farmers to respond by planting less.

There is a second effect from inventories. When a commodity is scarce, prices rise and the curve flips into backwardation, which signals scarcity to everyone. Scarcity does not resolve in a week. So both the price and the curve tend to keep pointing the same way for months.

Shorting the laggards is what makes this different from simply owning commodities. It strips out the overall commodity market direction and leaves you with a relative bet, which is much less correlated to inflation and to the general commodity cycle.

Where it works (regimes)

It works when different commodities are being driven by different fundamentals, which is most of the time. Energy is about OPEC and drilling, grains are about weather, metals are about Chinese construction. Those stories run on their own clocks, which is exactly what a relative strategy needs.

It works badly when everything in the complex is being driven by one macro factor. In a broad inflation shock or a global growth scare, all commodities move together, and the spread between the winners and the losers narrows. The strategy has nothing to work with, and it also tends to give back gains when the macro driver reverses.

The 2008 crash is the standard cautionary example: commodities had trended up together for years, the momentum book was long energy and grains, and the reversal was brutal and fast.

Signals

• The classic signal is the trailing 12-month return, ranked across the universe. Shorter windows of 3 to 6 months also work in commodities, arguably better than in equities, because the physical adjustment cycle is shorter than the corporate one.
• Use futures returns computed from a properly rolled series. Using spot prices is a common and serious error, because it ignores the roll cost, which in a contangoed market is most of what you actually earn or lose.
• Consider adding the curve slope to the ranking. Momentum and carry are quite complementary in commodities, and a market that is both trending up and in backwardation is a much stronger signal than either alone.
• Skip-month conventions matter less here than in equities, since the short-term reversal effect is weaker in futures.

Portfolio construction

The universe is small. You might have 25 markets, but they cluster into perhaps five real sectors, and within a sector they are highly correlated. Corn, wheat and soybeans are not three independent bets.

Cap risk by sector. Without that cap, a ranking will regularly leave you long every energy contract and short every grain, which is a single macro bet dressed up as a diversified portfolio.

Volatility-scale everything. Natural gas is many times more volatile than gold, and an equal-notional book is really just a natural gas fund.

Risk model

The tail is a sharp reversal, and it comes from the same place as the return: when the macro driver flips, all your longs fall and all your shorts rally at once.

Because the universe is small, the strategy cannot rely on diversification to smooth this. Run it at a modest volatility target and accept that the drawdowns will be lumpier than a broader multi-asset trend programme.

Watch for events that no price signal can anticipate: a weather shock, an export ban, an OPEC surprise. These invalidate the fundamental story overnight, and the signal has no way to know.

Costs & implementation

Costs vary enormously across the complex. Crude oil, gold and corn are cheap to trade. Coffee, cocoa, lean hogs and orange juice are not. If your backtest applies one cost number across all markets, it is wrong, and it is wrong in the direction that flatters you, because the extreme ranks tend to fall in the thin markets.

Rolls are a large part of the total cost. Since the strategy holds positions for months and the contracts expire, you are rolling regularly, and every roll is a trade.

Capacity is genuinely limited. The full commodity futures complex is small compared with equities or rates, and the thinner markets cannot absorb much. This is a strategy for a moderately sized book, not for a giant one.

Failure modes

• Correlated reversal when a macro driver flips.
• Sector concentration masquerading as diversification.
• Using spot prices instead of rolled futures returns and producing a backtest that cannot be traded.
• Underestimating costs in the softs and livestock markets.
• Assuming the strategy scales.

Our Notes & Suggestions

Blend momentum with the curve signal. The evidence that both work in commodities is reasonably strong, and their combination is more robust than either alone. A simple approach is to average the two ranks and trade the combined ranking.

Test the strategy with the thin markets removed. If the whole result comes from coffee and orange juice, you do not have a strategy, you have a data artefact.

Finally, remember what this strategy is for. Its value in a portfolio comes from being uncorrelated with equities and bonds, not from a spectacular standalone Sharpe ratio. Judge it on that basis and size it accordingly.