A similar chart does not mean the same object
CFTC glossary definitions help distinguish calendar, inter-commodity and processing spreads. Start by asking what should be neutralised: physical quantity, monetary value, approximate factor sensitivity or a specific business risk. These objectives produce different coefficients. “Balanced spread” has no single meaning without the criterion.
Even equal initial monetary values do not guarantee equal reaction to later price moves. State the desired neutrality and show both the theoretical ratio and the implementable whole-contract position.
Check units with fictional contracts
Suppose contract A contains 10 tonnes and contract B 25 tonnes. To compare 50 tonnes of each commodity, use five A contracts and two B contracts. Initial monetary values are 5,000 and 4,500. Physical equality is not monetary equality.
In the author’s example, long A rises from 100 to 103 while short B rises from 90 to 94. On 50 tonnes, A contributes +150 and B contributes −200, for a total of −50. Both prices rose, yet the relative position lost. All figures are fictional, costs are excluded and no existing exchange contract is described.
If the purpose was equal initial monetary value, another coefficient would be needed. It might not be an integer in contracts, and rounding would leave residual exposure. Do not hide that difference under “the market is neutral”.
Currency, quality and delivery location
A price per barrel cannot be compared with a price per tonne without a meaningful conversion, and one currency cannot be compared with another without an exchange rate and date. Physical conversion may require product-specific characteristics, not a universal constant from a random table. Delivery location, logistics, quality and demand calendar change the economic content. A material difference is not free profit.
In refined products, a processing relationship does not imply identical movement in crude input and finished output. EIA’s crack-spread material compares crude input with refined products and warns that revenues and costs are not fully captured. Economic connection creates a research question; it does not fix a permanent price coefficient.
Two venues add infrastructure risk
For different exchanges, check trading hours, settlement currency, delivery rules, collateral, intermediary access and the ability to roll or transfer the asset. Screen prices may refer to different moments. If one market is closed or stale, the calculated spread may not be executable. Include commissions, money transfers and restrictions on moving the commodity.
Two legs do not necessarily share collateral netting. A small relative-price risk can coexist with substantial cash needs on each side. If one infrastructure fails, the offsetting position may be temporarily unavailable. Correlated lines on a chart do not eliminate operational risk.
Historical correlation is not an obligation to revert
The relationship can change with new technology, logistics constraints, quality, delivery regime and participants. If the price ratio departs from its past average, it need not return within a convenient time. A historically stable coefficient still needs testing in the new regime and an economic explanation. Mean-reversion analysis is a separate question.
Keep primary prices, units, dates, coefficients, available volumes and exceptions in the research journal. Compare physical, monetary and risk-weighted alignment without choosing the best variant retrospectively. The conclusion should describe residual risks and conditions, not promise profit from a “too large” difference.
Changing the objective changes the coefficient
Finish the sentence: “These quantities were chosen to equalise …”. The answer must name a specific quantity and formula. If the coefficient is statistical, retain the estimation period and uncertainty rather than only a rounded number. If it is physical, state units and product characteristics. Coincident numbers do not make two different objectives the same task.
Sources and verification limits
[1] CFTC Futures Glossary: classification of calendar, inter-commodity and product spreads.
[2] EIA, “An introduction to crack spreads”: the difference between crude and a refined-product basket; not net refinery margin.
The original CME materials could not be obtained in full. For the stated definitions, complete primary substitutes listed above were used; only the named concepts and educational arithmetic were checked. Trading interpretations and predictive promises were not validated.
Three refined-product pairs answer different questions
A product’s connection with crude comes through refining, but the product has its own demand, inventories and availability. EIA describes a one-commodity crack as comparing a barrel of product with a barrel of feedstock: an approximate price relationship, not actual refinery profit. With fictional distillate at 95 and crude at 75, distillate minus crude is 20. At 105 and 78, it is 27. Check diesel demand, regional inventories and delivery too. “Heating oil” does not substitute for an arbitrary residual-fuel specification.
A second pair is gasoline versus crude. Travel, refining availability and seasonal gasoline characteristics can differ from heating demand. With fictional gasoline at 92 and crude at 75, the difference is 17; at 94 and 78, it falls to 16. A common rise in crude and product does not guarantee a wider relative difference.
A third pair is distillate versus gasoline, comparing two refinery outputs rather than input and output. At 95 and 92 the difference is 3; at 105 and 94 it is 11. A cold winter and strong driving season can change demand differently. Refinery output is constrained by technology and feedstock quality; equal mass or value does not make products instantly interchangeable. These are three different economic relationships, not three names for one margin.
Gold and silver: one metal category does not mean one demand
USGS describes monetary, investment, jewellery and industrial uses for gold, while silver has substantial electronics and photovoltaic uses. Investment demand and industrial output can therefore affect the metals differently. Scrap recycling and co-production with other metals also matter: higher silver demand does not immediately expand a silver-only mine.
In the author’s example, gold is 2,000 and silver 25 per troy ounce. Gold/silver is 80. At 2,100 and 30, it is 70. Gold rose 5% and silver 20%; a lower ratio does not mean gold fell. One ounce of gold equals 80 ounces of silver at the initial prices for an equal-money comparison. Long one ounce of gold and short 80 ounces of silver then produce 100 − 400 = −300 before costs. This is not a forecast or contract ratio.
Platinum versus gold and silver: different industrial loads
USGS describes platinum-group metals as used in automotive catalysts and other industrial processes, with supply affected by mining and recycling. Changes in vehicle production, catalyst composition, scrap availability or mining infrastructure can differ from investment demand for gold bars. Silver also has its own electronics and solar demand.
Suppose platinum, gold and silver initially cost 1,000, 2,000 and 25 per ounce. Platinum minus gold is −1,000, platinum/gold is 0.5 and platinum/silver is 40. At 1,100, 2,100 and 30, the absolute difference remains −1,000 while the ratios become approximately 0.5238 and 36.6667. An unchanged absolute gap and a changed relative valuation are compatible. The model explains the comparison, not a required return to a historical average.
Index versus index and index versus stock
Indexes differ in constituents, weights and review rules. FTSE Russell methodology separately defines size selection and growth/value segmentation; changes in rules or constituents change economic exposure. Differences in index levels with different bases say almost nothing. Compare aligned returns and notional leg values first, then state currency, dividends and derivative multipliers.
Fictional index exposures of 10,000 produce +800 on the long leg after an 8% rise and −500 on the short leg after a 5% rise, for +300 before costs. This does not neutralise sector, country or size factors. In stock minus index, company results, debt and corporate events remain. If the index rises 5% and stock 2%, equal initial exposures of 10,000 give −300. The stock is not thereby required to catch up.
A currency pair requires both currencies to be defined
A currency quote is a relative price. If A/B means units of B per A, A/B = 1.20 and C/B = 0.80 imply A/C = 1.50 units of C per A. You cannot subtract 1.20 and 0.80 and call the result A/C. At 1.26 and 0.84, the cross remains 1.50 even though both quotes rose 5%.
A BIS speech by Benoît Cœuré discusses currency dynamics through relative-rate expectations and capital flows and warns that a simple observed relationship can change. The rate on one currency alone does not determine the pair. Two currency futures also require aligned notionals and result currency. Different settlement currencies cannot be added without a stated rate.
WTI–Brent: quality, hub and route to the buyer
EIA’s historical example explains how delivery from an inland hub to the coast, Cushing inventories and export cost affected WTI relative to Brent. The mechanism is general, not a claim about current pipelines or historical tariffs. Different benchmarks are not necessarily interchangeable at one place and month.
With fictional WTI at 70 and Brent at 75, WTI minus Brent is −5. If inland export is constrained, prices may become 68 and 76, making the spread −8; the change of −3 is not explained only by the world oil price. New transport capacity, crude quality or destination can change the mechanism. Buying one futures contract and selling another does not automatically provide a right to move physical oil or net collateral across venues.
Heating oil–gasoil: do not subtract a gallon from a tonne
A U.S. distillate benchmark and European gasoil add quality, units, geography and currency at once. Read both specifications and delivery dates rather than treating “diesel” as full identity. Regional inventories and delivery can differ; EIA shows the role of heating, agricultural and freight demand in distillate balances.
For a purely educational conversion, suppose one tonne of the selected product contains 300 fictional volume units. This is invented and not a real gasoil conversion factor. At 2 per unit, the first benchmark is 600 per tonne; the second is 620, so the difference is −20. If the volume price becomes 2.10 and the second benchmark 625, the difference is +5. Actual density, measurement temperature or currency would change the calculation. None of these examples is a trading signal.