Contango and Backwardation: How to Read the Shape of a Futures Curve

# Contango and Backwardation: How to Read the Shape of a Futures Curve

The one takeaway: A futures curve prices storage, financing, and the value of holding the physical commodity now. Its shape describes market structure and the cost of rolling a position, not the direction the price is going.

A futures curve, also called the term structure, plots the settlement prices of the same commodity across a series of future delivery months, from the nearest contract to the furthest out. Two words describe the slope of that line. When later delivery months cost more than nearer ones, the curve is in contango. When later months cost less, the curve is in backwardation. This paper explains what each shape encodes and shows how to read the four commodity curves that Kresmion Research tracks, using the settlement of Friday 2026-07-17.

Key takeaways

What a futures curve actually is, and how to name its shape

Contango and backwardation describe the slope of the curve, not its level. A market can be in contango whether prices are high or low; the word is only about whether the next month costs more or less than the one in front of it.

The CME Group defines the two states directly: in contango the forward price of a futures contract is higher than the spot price, an upward sloping curve; in backwardation the forward price is lower than the spot price, a downward sloping or inverted curve (CME Group Institute).

Kresmion Research turns that idea into a fixed rule so the label is reproducible. It computes the front-to-second-month spread as (second_month_close minus front_close) divided by front_close. When that spread sits above +0.25% the curve is labeled contango; below -0.25% it is backwardation; anything in between is flat. One clarification matters for honesty: this rule compares two futures, the front month and the second month, not a future against spot. The CME definition above is stated against spot, and the front-month future is not spot; the two comparisons usually agree but can diverge as the front contract nears expiry. The plus or minus 0.25% band and the exact formula are published on the Kresmion methodology page.

The mechanics: cost of carry

The reason a deferred contract is priced above or below the near one comes down to the cost of carry. A future delivery price is roughly the spot price plus what it costs to carry the physical commodity forward to that date. Carry has three parts. Two push the curve up: financing, meaning the interest on cash tied up, and storage and insurance. One pushes it down: the convenience yield.

The CME notes that physically delivered futures "may be in a contango because of fundamental factors like storage, financing (cost to carry) and insurance costs," because supplying the commodity at a later date costs more (CME Group).

Convenience yield: the non-obvious term

Backwardation is the mirror image. When there is a real benefit to holding the physical commodity right now, near-dated contracts trade at a premium to deferred ones. The CME calls this the convenience yield: "a benefit to owning the physical material, such as keeping a production process running," an implied return on inventory that "is inversely related to inventory levels. When warehouse stocks are high, the convenience yield is low and when stocks are low, the yield is high" (CME Group).

Convenience yield is inferred from prices, not paid in cash. This inventory-driven view of the curve is the theory of storage, developed across the mid-twentieth century: Nicholas Kaldor introduced the convenience yield in 1939, and Holbrook Working formalized the supply-of-storage (or price-of-storage) relationship in the late 1940s. Ample supplies and high inventories tend toward contango, while tight supplies and low inventories tend toward backwardation (Theory of storage, Wikipedia). The inventory data that informs this view differs by market: the EIA publishes weekly stocks for crude oil (its Weekly Petroleum Status Report) and natural gas (its Weekly Natural Gas Storage Report), corn stocks come from the USDA, and gold warehouse stocks are reported by the COMEX exchange and the LBMA. Kresmion reports curve shape only; it does not ingest these inventory series or decompose carry into its component parts.

Reading the four Kresmion curves (as of Friday 2026-07-17)

Kresmion assembles curves for four physical-commodity roots, using roughly the front eight delivery months of daily settlement closes (on 2026-07-17 gold lists eight contracts and the other three list seven). On the futures-curve tool, the settlement of Friday 2026-07-17 shows two markets sloping down and two sloping up.

WTI crude and natural gas: backwardation

WTI crude oil is in backwardation. The nearest contract, CLQ26 (August 2026), settled at 82.49 and the next listed, CLU26 (September 2026), at 81.78, a front-to-next spread of -0.86%. The whole listed strip slopes down cleanly from 82.49 (August 2026) to 75.68 (February 2027), a decline of 8.26% across seven listed contracts. This is a textbook downward-sloping curve with no seasonal kinks (Kresmion futures-curve tool, as of 2026-07-17).

Natural gas shows a front-to-next spread of -1.20%, with NGQ26 (August 2026) at 2.911 versus NGU26 (September 2026) at 2.876. Here an honesty point matters: that label describes only the first two contracts. The full natural-gas curve is seasonal, not a clean downward slope. It rises steeply into winter to 4.160 at January 2027, roughly 43% above the August front, then eases to 3.765 at February 2027. The front is backwardated, but the shape overall is a winter heating premium, not a uniform inversion (Kresmion futures-curve tool, as of 2026-07-17).

Gold and corn: contango

Gold is in contango. GCQ26 (August 2026) settled at 4018.80 and the next listed, GCV26 (October 2026), at 4046.00, a front-to-next spread of +0.68%. The strip rises steadily to 4233.00 at October 2027, up 5.33% across the listed contracts, the classic upward carry slope. Two caveats. First, the front-to-next here spans a two-month gap, from August to October, because of the data source rather than a missing contract. COMEX gold's liquid cycle runs February, April, June, August, October, December, and the free feed used here lists only those months; the exchange also carries thinly traded serial contracts for nearer calendar dates (a September 2026 among them), but they are not in this source, so the contract that follows August in our data is October. Second, the deferred 2027 contracts are thin, trading roughly 5 to 65 lots on that date, so those far prices are settlement marks rather than actively traded (Kresmion futures-curve tool, as of 2026-07-17).

Corn is also in contango, with a front-to-next spread of +5.12%: ZCU26 (September 2026) at 444.75 versus ZCZ26 (December 2026) at 467.50. This is the least clean of the four. The front-to-next pair spans a three-month gap across the US harvest (corn lists on a March, May, July, September, December cycle), and the strip is not a simple upward slope: it rises to 496.75 at July 2027, then drops to 484.25 at September 2027, the classic old-crop to new-crop step-down. Corn's contango reflects storage cost plus harvest seasonality, not pure cost-of-carry (Kresmion futures-curve tool, as of 2026-07-17).

Because the front-to-next gap differs by commodity (crude and gas one month, gold two months, corn three months), the raw front-to-next percentages are not directly comparable across the four markets. That is why the next section reports a gap-normalized figure.

Rolling a position and roll yield

A futures contract expires. To keep exposure past expiry, a holder rolls: closing the expiring contract and opening a later-dated one. The shape of the curve decides whether that roll is a headwind or a tailwind.

In contango you sell the cheaper expiring contract and buy a pricier deferred one, a recurring drag known as negative roll yield. Fidelity illustrates the mechanic: rolling a 100 dollar contract into a 101 dollar contract forces you to buy nearly 1% less of the commodity, and "a 1% monthly cost comes to a nearly 13% cost on an annualized basis" (Fidelity). In backwardation the roll runs the other way, selling the richer near contract and buying a cheaper deferred one, a positive roll yield.

Kresmion annualizes this from the front-two spread as ((front_close divided by next_close) raised to the power (12 divided by month_gap), minus 1), times 100. Normalizing for the differing month gaps puts the four markets on one scale. As of Friday 2026-07-17 the tool reports: WTI +10.93%, natural gas +15.62% (both backwardation, reported positive), gold -3.97%, and corn -18.09% (both contango, reported negative). Read the seasonal markets with care. Natural gas's +15.62% is one shoulder-season month, August to September, extrapolated to a full year; a position actually rolling natural gas through the autumn into the steeply higher winter strip would face negative roll, not a positive tailwind. Corn's large negative figure is amplified by the harvest step-down and should be read with the seasonality caveat, not as a pure storage cost. Each value is a descriptive property of the front-two spread at that settlement extrapolated to twelve months, not a realized return and not a Kresmion Research forecast or recommendation.

Why the shape is not a forecast

It is tempting to read a rising curve as the market predicting higher prices. That reading is not reliable. The EIA describes backwardation as "higher prices for near-term contracts compared to prices for contracts with later delivery dates", which the agency explains as a curve that anticipates lower prices ahead (EIA), framing the slope as reflecting current anticipation. Kresmion Research treats that as one framing, not as fact, and reports the slope as structure rather than as a forecast; whether a curve's shape predicts future spot prices is a debated question, not a settled rule.

Two distinctions keep this honest. First, normal backwardation is a separate idea, associated with Keynes (A Treatise on Money, 1930) and later Hicks (Value and Capital, 1939), about futures sitting below the *expected future* spot price because hedgers pay a risk premium to speculators (Normal backwardation, Wikipedia). It is a claim about futures versus the *expected future* spot, not the same as simply observing a backwardated curve today, which compares futures to the *current* spot. Second, normal backwardation is a hypothesis about hedging pressure, not an established law, so it does not license reading today's curve slope as a forecast of where prices will go.

One more anchoring fact makes the roll matter: as a contract nears delivery, its price converges toward spot, "otherwise an arbitrage opportunity would exist" (CME Group). The gap between a deferred contract and spot is not free money; it closes as delivery approaches, and the curve's shape describes what a holder pays or earns as positions roll toward spot, if the shape holds.

Honest limitations

This monitor is deliberately narrow. Read the numbers with these constraints in mind.

• Coverage is four markets. Kresmion assembles curves for only WTI crude (CL), natural gas (NG), gold (GC), and corn (ZC), roughly the next eight delivery months each (seven or eight depending on the root), even though the underlying contract table lists 20 roots. There are no equity-index, Treasury, or other-commodity curves here.
• Two of the four curves are seasonal, not clean carry curves. Natural gas is labeled backwardation from the front two months only; its full curve is a winter premium. Corn's contango spans the US harvest and steps down from old crop to new crop. Only gold is a clean cost-of-carry example.
• The snapshot is two days old and a single point. The only settlement in the data is Friday 2026-07-17; the pull ran Sunday 2026-07-19, and there is no Saturday 2026-07-18 settlement. The spread history accrues daily and currently holds one point, so this table cannot show whether a curve's contango or backwardation deepened or eased. Any trend language about the shape is unsupported here.
• Front-to-next gaps are not uniform, so raw front-to-next percentages are not comparable across commodities. Use the annualized roll-yield field for cross-commodity comparison.
• The annualized roll yield extrapolates one month to a year. For a seasonal curve this can mislead. Natural gas's +15.62% comes from a single August-to-September shoulder-season spread, but a position rolling through the autumn into the steeply higher winter strip would face negative roll, not a positive tailwind. Read the annualized figure as a property of the front-two spread at one settlement, not a return any holder is guaranteed to earn.
• Imminent front-month expiries sometimes drop out of the free daily source, so occasionally the contract used as the front is really the second delivery month, which shifts both the spread and the roll-yield reading. Gold shows a related quirk: the free feed lists only gold's standard February, April, June, August, October, December cycle and omits the exchange's thinly traded serial months (including a September), so the gold front-to-next spans a two-month gap.
• Deferred liquidity is thin. Gold's 2027 contracts traded roughly 5 to 65 lots and corn's back months only a few thousand, so far-dated closes are settlement marks, not actively traded prices. Front-month liquidity is solid.
• The plus or minus 0.25% flat band is a presentation choice, not an economic constant. A mildly but genuinely sloped curve can read as flat, and any label near the threshold is sensitive to small daily noise.
• Convenience yield and carry components are inferred, not measured. The monitor reports curve shape only; it does not ingest inventory data (which for these markets would come from the EIA for crude oil and natural gas, the USDA for corn, and exchange or LBMA warehouse reports for gold) and it does not decompose carry into financing, storage, and convenience yield.
• The annualized roll-yield field is a Kresmion-computed derived metric returned by the platform; it is cited as the platform's official value. The raw front-to-next spreads were independently reproduced from the database closes and match to four decimals.

Methodology and sources box

• What is measured: four physical-commodity roots (CL, NG, GC, ZC), up to the front eight delivery-month settlement closes (seven or eight per root), the front-to-second-month spread, its state (contango, backwardation, or flat), and the annualized roll yield, refreshed daily.
• Thresholds and formulas: contango above +0.25%, backwardation below -0.25%, flat between; spread and roll-yield formulas as above. Published on the Kresmion methodology page.
• Data source: daily settlement closes from a free public source (Yahoo Finance), stored in the read-only production table `futures_curve` and served by the public endpoint used by the futures-curve tool. Both return identical closes and volumes for the 2026-07-17 settlement. Inventory data (EIA, USDA, COMEX, LBMA) is named for context only and is not ingested by the monitor.
• As-of: settlement Friday 2026-07-17; pulled Sunday 2026-07-19.
• Attribution: Kresmion Research. Information only. No buy or sell recommendation, no target price, no forecast.

FAQ

Does contango mean prices are going to rise (and backwardation that they will fall)?

No. Contango and backwardation describe the slope of the curve, driven by carry: financing, storage, and the convenience yield of holding the commodity now. The textbook definition compares a futures price to spot, but Kresmion's own label is set from the front-month versus second-month futures spread, not from spot; the two usually agree and can diverge only as the front contract nears expiry. Either way the shape is not a prediction of the future price. Some agencies, including the EIA, describe the slope as reflecting what participants currently anticipate, but whether a curve's slope predicts future prices is a debated question, not a settled rule. Kresmion Research reports the shape as market structure, not direction.

What is roll yield, and why do the four commodities show such different numbers?

Roll yield is the effect of moving exposure from an expiring contract into a later-dated one. In contango you roll into a pricier contract, a drag; in backwardation you roll into a cheaper one, a tailwind. Kresmion annualizes it from the front-two spread, normalizing for the differing month gaps (crude and gas one month, gold two, corn three) so the four markets sit on one scale. As of 2026-07-17 the tool shows WTI +10.93%, natural gas +15.62%, gold -3.97%, and corn -18.09%. Treat the seasonal markets with care: natural gas's +15.62% comes from a single shoulder-season month, and a position rolling through the autumn into the higher winter strip would face negative roll instead; corn's -18.09% is inflated by the old-crop to new-crop harvest step-down. Each figure describes the front-two spread at one settlement extrapolated to a year, not a realized return.

Why is natural gas called backwardation if its curve rises into winter?

Because Kresmion's headline state is decided on the front two delivery months only. On 2026-07-17 the August-to-September spread is -1.20%, which is backwardation by the rule. But the full strip rises to 4.160 at January 2027, roughly 43% above the August front, then falls to 3.765 at February 2027. That is a seasonal winter heating premium, not a uniform downward slope, which is why this paper flags natural gas rather than presenting it as clean backwardation.

Sources

• CME Group Institute, "What is contango and backwardation?" (definitions, cost of carry, convenience yield, convergence): https://www.cmegroup.com/education/courses/introduction-to-ferrous-metals/what-is-contango-and-backwardation
• US Energy Information Administration (EIA), Today in Energy, 20 September 2013 (curve-shape definitions, WTI example): https://www.eia.gov/todayinenergy/detail.php?id=13051
• Theory of storage (Kaldor 1939, convenience yield; Working, supply-of-storage, late 1940s), Wikipedia: https://en.wikipedia.org/wiki/Theory_of_storage
• Normal backwardation (Keynes, A Treatise on Money, 1930; Hicks, Value and Capital, 1939), Wikipedia: https://en.wikipedia.org/wiki/Normal_backwardation
• Fidelity, "Commodity ETFs, contango and backwardation" (roll yield): https://www.fidelity.com/learning-center/investment-products/etf/commodity-etfs-contango-backwardation
• Kresmion futures-curve tool (live curve data, as of 2026-07-17): https://kresmion.com/tools/futures-curve
• Kresmion methodology (thresholds and roll-yield formula): https://kresmion.com/about/methodology