Continuous contract

How it works

There is no instrument called "crude oil" or "the DAX future". There is a chain of separately listed contracts, each with its own delivery month, its own expiry rule and its own order book. ICE Futures Europe lists Brent Crude futures for up to 156 consecutive months, with trading in each one ceasing at the end of the settlement period on the last business day of the second month before the contract month. Eurex lists DAX futures on the March, June, September and December cycle, with the last trading day the third Friday of the maturity month. A ten-year chart of either market is therefore not a record of one thing trading; it is a splice of dozens.

Building the splice takes two decisions, and they are independent of each other.

The roll rule — when to switch. The common rules are a fixed calendar offset (roll N business days before expiry, or on the first day of the delivery month), a volume crossover (roll when the next month trades more than the front month), an open-interest crossover, or the Nth nearest contract held at a constant distance from expiry. A crossover rule does not fire on the same date every year, so it does not produce a series anybody else can regenerate without being told the threshold.

The adjustment method — what to do about the gap. At the roll the two contracts are at different prices, and that difference is not a return anybody earned. The options are to leave it alone (unadjusted or spliced, which puts a fictional jump in the series at every roll), to subtract the gap from all prior prices (difference or back-adjustment, sometimes called the Panama method), or to scale all prior prices by the ratio of the two (proportional or ratio adjustment). Forward-adjusted variants apply the same arithmetic in the other direction, holding the old prices fixed and moving the recent ones.

The prices in an adjusted series are not prices

This is the part that is easy to read past. A difference-adjusted series preserves the size of every daily move in currency terms but destroys the price level: after enough rolls the historical prices can be nothing like what traded, and in a market whose front month once settled below zero the adjusted history can contain negative numbers where none were printed. A ratio-adjusted series preserves percentage moves and destroys the absolute differences instead. An unadjusted series preserves both properties within each contract and inserts a fake return at every roll, which is the one that quietly poisons any statistic computed over the whole series.

So a level in an adjusted series cannot be compared with anything outside that series — not a strike price, not a round number somebody watched, not a quote from a news story, not the same market from another vendor built a different way. The same arithmetic applied to equities is in why adjusted close differs between sources, where the roll gap is replaced by a dividend or a split and the choice between difference and ratio is the same choice.

Why it matters here

Every product in futures data and futures trading tools that shows a long-horizon futures chart is showing a construction, and the construction is rarely on the front page. CSI Data builds eight of them and lets the roll timing be specified by calendar date, by days relative to the delivery month or by thresholds on volume and open interest; Norgate Data ships unadjusted and back-adjusted series side by side, which is the shape that lets you check one against the other. Backtrader has continuous cash adjustment for futures-like instruments in its simulated broker, and QuantRocket and AmiBroker each expect a series that was built somewhere upstream.

Two questions settle it before any result is read: which roll rule made this file, and which adjustment method. A source that answers neither cannot be reproduced and cannot be reconciled with a second source, which is the same failure point-in-time data addresses for fundamentals — the number is not wrong so much as undated and unattributable.