Module 8 of 14 · Directional strategies
Carry, the curve and rolling
A futures contract converges to the spot price as it approaches expiry. If later-dated contracts are cheaper than the one you hold — backwardation — a long position tends to drift upward as it ages even if spot never moves. If later contracts are more expensive — contango — longs bleed and shorts collect. That expected drift is carry, and you can read it straight off the curve without any forecasting model.
To turn it into a forecast: take the price difference between two contracts, divide by the near price, annualise by the time between expiries, then divide by volatility so that 5% carry in a calm bond future counts for more than 5% in a wild energy contract. Scale so a typical reading is 10 and cap at ±20. Seasonal markets need care: smooth the carry signal over weeks or months, otherwise you will trade the seasons rather than the carry.
Carry is the natural partner to trend. They are almost uncorrelated, and carry earns steadily in quiet markets where trend struggles. It also has the opposite skew: carry earns small and steady and occasionally loses hard when the curve flips.
Contracts and rolling — the plumbing
- Which contract? Only liquid ones. Many financial and metal futures are only liquid in the front month. Where several months trade well, avoid the front — it is jumpier and gets close to delivery. In seasonal commodities, holding the same calendar month each year (for example December) keeps seasonality out of your price series.
- When to roll? A few days before expiry or, for physically delivered contracts, before first notice day. Brokers will close you out — on their terms — if you forget.
- How to roll? As a single calendar-spread order rather than two separate legs: one spread instead of two, and no risk of one leg filling without the other. Better still, roll passively by placing your next signal trade in the new contract.
- Price history: splice contracts into a back-adjusted series (shift the history by the price gap at each roll) for trend signals, but size positions from the price of the contract you actually hold.
🎛 Curve carry lab
Curve shape
backwardation
Annualised carry
14.4%
70 vs 69.16
Risk-adjusted
0.48
carry ÷ volatility
Carry forecast
13.7
scalar 28.6
The curve here is a smooth illustration: price of the k-th contract = front × (1 + slope)^k. Carry compares the contract you hold with a later one: if later contracts are cheaper (backwardation) a long position drifts up toward spot as it ages and earns that difference. Annualise it, divide by volatility to put it in risk units, then scale by whatever makes a typical reading equal 10 — set “typical” from your own history. Capped at ±20.
Carry is mostly free information
Most of what carry tells you is already in the curve today; it needs no lookback window and changes slowly, so its turnover is low and its costs are small. The main design choices are which two contracts to compare (adjacent liquid ones, ideally the one you hold and the next one out) and how much to smooth.
A weakness worth knowing: carry is an expected drift, and spot price moves dwarf it day to day. A large backwardation in a market that is collapsing will still lose money. That is exactly why carry is combined with trend rather than traded alone.
Things to try
- • Flip the slope from −1.2% to +1.2% per month: a contango of the same size gives a short forecast of the same strength.
- • Raise volatility from 30% to 70% with the same curve: the forecast shrinks, because the same carry is worth less when the market moves more.
- • Compare the contract 1 month out with 6 months out on the same curve: annualising makes them consistent, but on a real curve the shape bends, so the choice of comparison contract matters.
Original interactive lesson. Labs run on a seeded synthetic market in your browser, built to illustrate mechanics, not to predict any real market. Contract figures are illustrative; check exchange specifications. Educational analysis, not investment advice.