Separate a futures roll from the return earned through time

Write the transactions rather than only the spread

A hypothetical long investor closes a near contract and opens a farther contract. Those are two transactions in different instruments. Futures margin is not payment of the full commodity value, so the difference between quoted contract prices should not be treated as a cash purchase invoice unless the accounting explicitly models physical ownership.

The CFTC’s commodity ETP advisory explains why a futures-based vehicle can behave differently from the underlying commodity’s cash price. This publication uses that distinction to motivate careful accounting. It does not calculate an actual fund return, recommend a roll schedule or assume that a particular curve shape persists.

Follow an invented position across the roll

Suppose a one-unit illustration closes the near future at 80 and opens the far future at 82. The two-unit price gap describes the replacement exposure at that instant. It is not an automatic two-unit loss posted merely because the new contract has a higher quoted price.

If the far contract subsequently falls from 82 to 81 while held, its price-change profit or loss is minus one per unit before costs. If it rises to 84, the change is plus two. Those outcomes require subsequent observations. They cannot be known from the original 80/82 snapshot or its annualized slope alone.

Specify what is held constant

Keeping one contract maintains the specified physical quantity but changes price-based notional when the replacement contract is differently priced. Holding a fixed notional instead changes the number of contracts, often requiring rounding. A study must state which exposure rule it uses before comparing returns across roll dates.

For a synthetic notional of 8,200 and a one-unit contract, price 82 corresponds to one hundred units. Price 80 would correspond to 102.5 units before integer constraints. That difference illustrates why a constant-unit calculation and a constant-notional index are not interchangeable. Real contracts have their own multipliers and indivisibility.

Avoid double counting a roll component

A decomposition into spot movement, roll effects and collateral return depends on the chosen index or accounting definition. If a continuous series already embeds contract transitions, adding an independently calculated spread term may count an effect twice. Inspect how the source constructs the series and which prices are actually held through time.

Back-adjusted price histories can be useful analytical references but can change historical levels. They are not automatically executable return histories. A reliable performance study would retain individual contract identities, transition dates, quantities and costs. The lab deliberately stops at a two-price diagnostic rather than fabricating that missing path.

Keep collateral and implementation visible

Cash collateral can earn or incur a separate return, while margin changes can create liquidity needs. Commissions, bid-ask spreads and order timing also affect realized results. None is captured by the absolute far-minus-near spread. An unqualified statement that contango costs a fixed percentage omits the exposure and time conventions needed to evaluate it.

The worked example is an accounting explanation, not a backtest. A research dossier should distinguish its static curve calculation from any separately evidenced investment return. This makes the result useful without pretending that observing two contract prices is enough to reconstruct a tradable strategy or an investor’s total experience.

Keep in the decision record

• Closing one contract and opening another is not an instantaneous spread return.
• State whether units or notional are held constant.
• Do not add a roll term to a series without understanding its construction.

Source notes

1. CFTC — Risks Before Investing in Commodity ETPs or Funds
2. CFTC — Economic Purpose of Futures Markets and How They Work
3. CME Group — What Is Contango and Backwardation, official education

Source dates and scope matter. Historical inventory observations are retained snapshots; forecasts and synthetic curve assumptions are labelled separately. No investment outcome or product coverage is implied.

Continue the analysis