Continuous Futures Contract Construction

A futures contract expires — a March crude oil contract stops trading and settles in March, and anyone still wanting exposure has to hold a completely different contract, June crude oil, which trades at its own separate price. No single futures contract has a multi-year price history the way a stock does. To backtest a futures strategy over years of data, you need to stitch together a sequence of individual contracts into one continuous series, and the specific way you do that stitching — which contract you're "in" on any given date, and how you handle the price gap at each switch — can change a backtest's results substantially.

The idea

Two choices dominate continuous contract construction. First, the roll rule: on what date do you switch from the expiring contract to the next one — a fixed number of days before expiration, or the date volume shifts to the new contract, are both common conventions, and they don't always agree. Second, the price adjustment method at each roll: because the expiring and the next contract usually trade at different prices (reflecting time value, storage costs, or seasonal expectations), simply concatenating raw prices creates an artificial jump at every roll date that doesn't represent any real gain or loss to a holder who rolled their position forward. The common fix, "back-adjustment," shifts all prices before the roll by the difference (or ratio) between the two contracts' prices at the roll date, so the stitched series shows a smooth transition — at the cost of the resulting prices no longer being literal historical trade prices.

Worked example

A trader rolling a long crude oil position from the expiring March contract (trading at $78) into the June contract (trading at $79, reflecting the cost of carrying oil an extra three months) experiences no real gain or loss from the roll itself — they've simply exchanged one contract for an economically equivalent one at the prevailing spread. A raw, unadjusted continuous series that just switches from showing March's $78 to June's $79 would show a fake one-day $1 gain. A back-adjusted continuous series instead shifts every price before the roll date by that $1 difference, so the stitched series shows a flat transition, matching what the trader actually experienced.

What this means in practice

Different roll rules and adjustment conventions can produce meaningfully different backtested Sharpe ratios for the same underlying strategy, especially for strategies sensitive to short-term price moves around roll dates, which is why futures research reports typically need to state explicitly which roll and adjustment convention was used, and ideally test sensitivity to a few reasonable alternatives.