Continuous Futures Charts: Why Price Changes Are Not Trading Returns

In the teaching example below, a continuous futures chart rises from 98 to 102, a gain of 4.08%. A trader who holds ten long contracts across the same two delivery months earns exactly zero from the futures positions. Neither result needs a fee adjustment to explain it: the chart changes contracts halfway through.

To choose data for analysis or a backtest, start with the question. A continuous series links different contract months into a price history. A trading ledger records the contracts actually held, their quantities and their cash gains or losses. A smoother chart does not, by itself, become an investment return series.

Illustrative contract prices and chart adjustments; these are not market quotations.

Start with the cash ledger: ten contracts, no futures profit

Consider a hypothetical commodity. Contract A expires sooner than contract B; both are quoted in US dollars per unit, with 1,000 units per contract. Initially A trades at 98. At the roll, A is 100 and B is 104 at the same observation time. Later, B is closed at 102. We assume consistent price sampling and, separately, that each stated execution price is obtainable. A published settlement price alone would not establish that an actual order could fill there.

The base example excludes commissions, the bid–ask spread, slippage, cash interest and funding constraints. It does not remove the four-point calendar spread between A and B: that difference is an essential input.

Futures are generally marked to market in cash each day. The entry-to-exit calculations aggregate those gains and losses; they do not imply that settlement waits until a position is closed. A new B position opened and valued at 104 starts with zero P&L. Buying B at 104 after selling A at 100 does not create a separate $40,000 cash loss. Deducting that amount after the two holding-period calculations would count a loss that this ledger never incurred.

Nor does opening futures require paying the entire commodity notional. Nevertheless, keeping ten contracts changes notional exposure from 100 × 10 × 1,000 = $1,000,000 immediately before the roll to $1,040,000 after it. That 4% increase does not establish a matching 4% increase in margin requirements or overall risk.

The higher price of B is not a guarantee of a losing investment. If B instead finishes at 108, with all other assumptions unchanged, its holding-period gain is $40,000 and the combined futures profit is $60,000. The subsequent contract price and position size determine the result; the initial calendar spread alone does not.

Why the deferred contract costs more is a question about futures basis and term structure. Whether holding it makes money is a separate calculation using the actual contracts and their subsequent price paths.

Three histories from exactly the same quotations

A raw splice takes the selected price of A and then switches to B. A back-adjusted chart rewrites earlier prices to remove the discontinuity. The following two adjustment methods preserve B's latest, unadjusted prices and change only the earlier A segment.

The two roll rows describe different contracts at the same time, not a market that first reached 100 and then rallied to 104. On the raw splice, 102 ÷ 98 − 1 produces 4.08%, but includes a four-point jump that was not earned by holding either contract.

Additive adjustment: the roll difference is 104 − 100 = 4. Add four to A's old prices: 98 becomes 102 and 100 becomes 104. If B remains anchored at 104, subtracting four from old A would turn 100 into 96 and widen the gap to eight points—the wrong direction.

This preserves A's two-point rise, but not its percentage return. The real contract gained 100 ÷ 98 − 1 = 2.04%; the adjusted segment gains 104 ÷ 102 − 1 = 1.96%.

Ratio adjustment: multiply A's history by 104 ÷ 100 = 1.04. Its first value becomes 101.92. The segment retains its 2.04% change, but the point move is now 2.08 rather than two. The entire adjusted series gains 102 ÷ 101.92 − 1 = approximately 0.0785%.

These preservation claims apply to observations receiving the same adjustment factor, not automatically to every indicator spanning several rolls. Terminology also varies between data systems. Establish which end remains anchored, which history changes and how the factors are applied; a label such as “adjusted” is not a sufficient specification.

Changing the contract count changes the answer again

Now use a different position rule: establish $980,000 of notional exposure at each entry, leave the quantity unchanged within each holding period, and do not add A's profit to B's target notional. The $980,000 is also the separately chosen initial capital allocated to this example. It is neither an upfront commodity purchase payment nor the required margin.

A still requires $980,000 ÷ (98 × 1,000) = 10 contracts. B requires $980,000 ÷ (104 × 1,000) = 9.423076923 contracts. A earns $20,000; B loses $18,846.15. Combined futures profit is $1,153.85, or 0.1177% of the initial $980,000 allocation, using unrounded inputs.

That is neither the fixed-ten-contract result of zero nor the ratio chart's 0.0785%. Fractional contracts are allowed here only to isolate the mathematics. A live implementation must address whole-contract sizes, smaller contracts where available, residual cash, margin and execution differences.

A third rule explains why the ratio chart can sometimes match a strategy. If A's $20,000 gain is added to the new notional target, B starts at $1,000,000 of exposure, or 9.615384615 contracts. B then loses $19,230.77; total profit is $769.23, approximately 0.0785% of $980,000. The match depends on this single simultaneous roll, positive prices, resetting exposure to then-current equity, fractional quantities and zero costs and interest. It does not make ratio adjustment a universal account-return formula.

A price series, an investment index and an account are different objects

An investment index has position and weighting rules, not just a way of joining prices. CME's rolling futures methodology specifies contract units and roll weights; quantities can be reset during the roll. One old contract need not be replaced by one new contract. Its published index also excludes transaction and access costs, so it is not an actual investor's net-of-cost account return.

The S&P GSCI family distinguishes price, excess-return and total-return versions. Its price version reflects specified futures price levels, despite the “Spot” label; it is not a direct quotation for an immediately deliverable physical basket, nor a promise that no rolling occurs. Excess return applies contract returns and weights. Total return adds interest on the methodology's assumed investment cash. “Excess” describes that calculation, not promised outperformance. Other index families require their own methodology checks.

For an account, inspect the actual cash records. Suppose the fixed-ten-contract strategy has zero cumulative futures P&L, $1,200 of genuinely eligible cash interest and $300 in all relevant trading costs. Net profit is $900. Do not assume that the entire notional earns deposit interest, and do not subtract the four-point calendar spread again. An initial-capital denominator gives a simple period return only under the stated no-external-cash-flow assumption; deposits and withdrawals require a defined return convention.

Use the chart for signals, the original contract for execution

A separate example concerns price coordinates only: after A has risen to 100, a trader sets a long-position stop at 99. If that history is later shifted upward by four, the chart coordinate becomes 103. The historical order remains 99. This is not an initial stop above the original 98 entry, and no stop execution is assumed.

A backtest must map an adjusted signal to the original contract month and price. Keep contract identity, trade direction, timestamps, execution prices, multiplier, quantities, costs and adjustment factors alongside the information available when each decision was made. A smoothed historical value is not an executable historical quotation.

Contract selection has a clock, too. Using a day's final trading volume to choose the active contract at that day's opening introduces information that was not yet available. Later historical adjustments are a separate issue: they may change fixed thresholds or percentage calculations, while some same-segment price differences remain unchanged. Not every adjusted-data test is automatically forward-looking.

Zero and negative prices need explicit rules. A ratio with a zero denominator is undefined; a negative price has no real-valued logarithm, and returns across zero can lose their usual interpretation. An adjusted negative history may also differ from a genuinely negative original contract quote. Check the original month and the product's rules rather than assuming every futures contract permits negative prices.

The practical choice is therefore not simply “raw or smooth.” First specify the decision, then select the price history, and keep a separate executable-contract ledger. The method of a particular FastBull series must be checked in its data specification; none of the illustrative constructions above establishes which algorithm it uses.