Crack and Crush Spreads in Depth: Refinery and Soybean Processing Margins in Futures

A crack spread is the futures-market value of a refinery’s gross margin: the price of refined products (gasoline and diesel) minus the price of the crude oil they are made from, in dollars per barrel. A crush spread is the same idea for soybean processors: the value of the soybean meal and oil from one bushel minus the price of the bushel. Both are traded as combinations of separate futures contracts, both are used by processors to lock in margins, and both are margined more cheaply than their legs because the legs offset.

In short: the NYMEX 3-2-1 crack is [2 x RBOB x 42 + 1 x ULSD x 42 minus 3 x WTI] / 3, using 2 RBOB gasoline contracts, 1 NY Harbor ULSD contract and 3 WTI crude contracts. The CBOT board crush is meal ($/short ton) x 0.022 + oil ($/lb) x 11 minus soybeans ($/bu), traded as 10 soybean, 11 meal and 9 oil contracts. These formulas are approximations of real processing margins: actual refinery yields, crude grades, locations, extraction rates and operating costs differ. The spreads move with inventories, seasons and policy, and they have their own risks, including going negative.

The basic arithmetic is introduced in futures spread trading. This piece goes further: contract-level mechanics, how far the benchmark is from a real plant, what the official data says about recent margins, and how exchanges treat these spreads for margin.

Part 1: the crack spread

The three legs

CME Group’s handbook “An Introduction to Crack Spreads” lists the three contracts used to build NYMEX cracks:

	Light Sweet Crude Oil (CL)	RBOB Gasoline (RB)	NY Harbor ULSD (HO)
Contract unit	1,000 barrels	42,000 gallons	42,000 gallons
Price quotation	Dollars per barrel	Dollars per gallon	Dollars per gallon
Minimum tick	$0.01	$0.0001	$0.0001
Last trade	3 business days before the 25th of the prior month (4 if the 25th is not a business day)	Last business day of the prior month	Last business day of the prior month
Settlement	Physical	Physical	Physical

Two details matter. First, 42,000 gallons is exactly 1,000 barrels, so one contract of each leg is the same volume. Second, the products are quoted per gallon and crude per barrel, so product prices are multiplied by 42 before subtracting. The HO code is a holdover: according to the US Energy Information Administration (EIA), the contract’s underlying has been ultra-low sulfur diesel since spring 2013, and RBOB (reformulated blendstock for oxygenate blending) has been the gasoline underlying since 2006. More on the crude leg is in crude oil futures explained.

Why 3-2-1

CME’s handbook explains that “a typical refinery adopts a 3:2:1 ratio as gasoline output is approximately double that of distillate fuel oil.” Other ratios exist for other plants: 1:1 (one product against crude), 5:3:2 for refiners with a higher distillate yield, and 2:1:1.

Terminology trips people up. In CME’s usage, a refiner who wants to protect against a shrinking margin sells the crack: it sells product futures and buys crude futures. The direction refers to the product side. Buying the crack is the opposite position, and it gains when products rise relative to crude.

Worked example: per barrel and per contract set

Illustrative prices for the same delivery month:

WTI crude: $75.00 per barrel
RBOB: $2.2000 per gallon, or 2.20 x 42 = $92.40 per barrel
ULSD: $2.5000 per gallon, or 2.50 x 42 = $105.00 per barrel

The two simple cracks:

Gasoline crack: $92.40 minus $75.00 = $17.40
Diesel crack: $105.00 minus $75.00 = $30.00

The 3-2-1:

[2 x 92.40 + 1 x 105.00 minus 3 x 75.00] / 3 = 64.80 / 3 = $21.60 per barrel

Per contract set (3 CL, 2 RB, 1 HO), the spread covers 3,000 barrels of crude, so its value is $21.60 x 3,000 = $64,800.

Sensitivities a desk checks first:

A 1 cent per gallon move in RBOB is $420 per RB contract (0.01 x 42,000), and moves the 3-2-1 by 2 x 0.42 / 3 = $0.28 per barrel.
A $1 per barrel move in crude is $1,000 per CL contract and moves the 3-2-1 by exactly $1 in the opposite direction.
Gross notionals are large even though the spread is small: about $92,400 per RB, $105,000 per HO and $75,000 per CL at these prices, or $514,800 gross across the six contracts for a position whose value is $64,800.

If the refiner sells the crack at $21.60 and the crack falls to $15.00 by the time it runs the crude, the futures gain is (21.60 minus 15.00) x 3,000 = $19,800, offsetting the lower cash margin. CME’s handbook works the same mechanics with its own example: a refiner sells the 3-2-1 at $30.40, the cash margin falls to $24.60, the futures gain is $5.80, and the hedged margin is the original $30.40. This is a classic short hedge; the general mechanics, including basis risk, are in how hedging with futures works.

How far 3-2-1 is from a real refinery

The benchmark assumes every barrel of crude becomes two-thirds gasoline and one-third distillate, with nothing else. The EIA’s own answer to “How many gallons of gasoline and diesel fuel are made from one barrel of oil?” is that US refineries produce “about 19 to 20 gallons of motor gasoline and 11 to 13 gallons of ultra-low sulfur distillate fuel oil” from a 42-gallon barrel. The 3-2-1 implies 28 gallons and 14 gallons.

Run the illustrative prices through EIA’s midpoint yields (19.5 gallons of gasoline, 12 gallons of distillate):

Gasoline: 19.5 x $2.20 = $42.90
Distillate: 12 x $2.50 = $30.00
Total: $72.90, from a barrel of crude that cost $75.00

Those two products alone do not cover the crude. The rest of the barrel (jet fuel, propane, petrochemical feedstocks, asphalt, residual fuel, petroleum coke) and the refinery’s processing gain must cover it, and some of those products sell below crude. A real refiner’s margin also depends on:

Crude slate and location. A Gulf Coast plant running heavy sour crude does not pay WTI Cushing prices, and its product sells at Gulf Coast or Midwest prices, not New York Harbor.
Timing. CME’s handbook notes that refiners typically buy crude for processing in one month and sell products one month later, so the hedge often pairs different contract months.
Compliance and operating costs. Fuel standards, renewable fuel obligations, energy and labour costs come out of the gross margin before anyone earns a return.

So the 3-2-1 is best read as a directional indicator of refining economics and a liquid hedge, not a profit-and-loss statement.

What the official data shows

The EIA publishes daily prices that let anyone rebuild the crack. Its front-month NYMEX futures series for crude, RBOB and ULSD run in its downloadable files from October 2005 to April 5, 2024. Its spot series for WTI Cushing, New York Harbor conventional gasoline and New York Harbor ULSD run to the present (the ULSD series starts June 14, 2006). We computed both in Python.

From EIA front-month futures (3-2-1, annual average, dollars per barrel):

Year	3-2-1	Year	3-2-1
2008	9.93	2020	11.27
2012	30.03	2021	19.68
2016	14.93	2022	38.47
2019	18.45	2023	32.09

The 2011-2023 average was $22.28. The highest monthly average in that series was $57.55 in June 2022. The lowest was $3.14 in November 2008, when the gasoline crack on its own averaged minus $5.44 for the month: gasoline futures traded below crude. A refiner who had bought the crack expecting mean reversion would have watched it go through zero.

From EIA spot prices (New York Harbor gasoline and ULSD against WTI Cushing), 2026 has been extreme. The monthly average 3-2-1 was $28.78 in February 2026, $46.65 in March, $65.62 in August and $68.22 in September 2026, the highest monthly average since the ULSD series began in 2006. In September, WTI averaged $97.31, gasoline $3.434 per gallon and ULSD $4.956. By November 3, 2026, the spot 3-2-1 was $44.27 with WTI at $99.64. The diesel side has been doing most of the work, and EIA’s weekly inventory data is consistent with that: distillate stocks were 99.3 million barrels in the week ending October 30, 2026, against an average of 130.1 million barrels for October weeks in 2015-2025, while refineries ran at 94.5% of capacity. EIA’s data shows what happened, not why, and we do not attribute the 2026 move to a specific cause here.

Listed crack products and margin credits

Most crack trading is done by combining the CL, RB and HO legs, and CME’s materials treat the 3-2-1 that way. NYMEX also lists financially settled crack spread contracts, so a trader can hold the spread as a single position. For example, NYMEX’s June 15, 2011 filing to the CFTC (Submission 11-226) self-certified the RBOB Gasoline vs. Brent Crack Spread Swap Futures (code RBB, Rulebook Chapter 1096): 1,000 barrels per contract, a tick of $0.001 worth $1.00, cash settled to a floating price based on the average of first nearby RBOB settlement prices minus the average of first nearby ICE Brent settlements.

On margin, the exchange recognises that the legs offset. CME’s handbook notes that traders benefit from “low margins due to the substantial spread credit received for margining purposes.” Its 2024 article “Trading Crack Spreads” puts a number on it: crack spreads “may require up to 30% less capital than an outright ULSD or RBOB futures position,” and a worked Micro 3-2-1 (two Micro RBOB, one Micro ULSD, three Micro WTI) needed $1,451 of margin, “a reduction of 26% from the outright Refined products futures margin requirements.” Those figures were as of that article in 2024; spread credits are set by the clearing house’s risk model and change with volatility. How the credits are calculated is covered in SPAN and futures margin models.

Part 2: the soybean crush

The yields and the formula

Contract background for the grains and oilseeds is in agricultural futures explained. CME Group’s “Soybean Crush Reference Guide” states that a 60-pound bushel of soybeans typically yields 11 pounds of soybean oil, 44 pounds of 48% protein meal, 4 pounds of hulls and 1 pound of waste. The three CBOT contracts:

	Soybeans (ZS)	Soybean Meal (ZM)	Soybean Oil (ZL)
Contract unit	5,000 bushels	100 short tons	60,000 pounds
Quoted in	Cents per bushel	Dollars per short ton	Cents per pound

Converting each to dollars per bushel:

Meal: price per short ton x 0.022 (44 pounds / 2,000 pounds)
Oil: price per pound x 11

Board crush ($/bu) = meal x 0.022 + oil x 11 minus soybeans

(If oil is quoted in cents per pound, use 0.11.) CME calls the cash-market version the gross processing margin (GPM) and the futures version the board crush. A processor hedging its margin “puts on the crush”: buys soybean futures and sells meal and oil futures. The reverse crush is the opposite.

Why 10, 11 and 9

The 1:1:1 ratio does not match the yields, so the exchange-defined board crush uses 10 soybean, 11 meal and 9 oil contracts. Ten soybean contracts are 50,000 bushels:

Meal: 50,000 x 44 lb = 2,200,000 lb = 1,100 short tons = 11 contracts (CME’s guide prints 1,110 tons; the arithmetic gives 1,100)
Oil: 50,000 x 11 lb = 550,000 lb = 9.17 contracts, rounded to 9

CME’s guide notes in a footnote that the standard unit “results in under-hedging the soybean oil component of the crush by 10,000 lbs.” The board crush trades on CME Globex as an exchange-defined spread (commodity code SOM), with implied matching against the outright legs.

Worked example with USDA’s forecasts

USDA’s Economic Research Service, in its Oil Crops Outlook of August 14, 2026, forecast season-average prices for the 2026/27 marketing year of $11.40 per bushel for soybeans, $310 per short ton for meal and 70 cents per pound for oil. These are annual averages, not futures prices, but they give a realistic, sourced set of inputs.

Meal: 310 x 0.022 = $6.82 per bushel
Oil: 0.70 x 11 = $7.70 per bushel
Board crush: 6.82 + 7.70 minus 11.40 = $3.12 per bushel

Per 10:11:9 contract set:

Meal: 11 x 100 x $310 = $341,000
Oil: 9 x 60,000 x $0.70 = $378,000
Soybeans: 10 x 5,000 x $11.40 = $570,000
Net: $149,000, or $2.98 per bushel

The per-bushel formula times 50,000 bushels gives $156,000. The $7,000 difference is exactly the 10,000 pounds of oil the 9-contract leg leaves unhedged (10,000 x $0.70). A processor hedging a real plant sizes the oil leg to its actual output, not the standard unit.

Sensitivities per set: 1 cent per bushel in soybeans is $500, $1 per ton in meal is $1,100, and 1 cent per pound in oil is $5,400.

In CME’s 2020 guide, the illustrative prices (meal $304 per ton, oil $0.3358 per pound, beans $9.44) give meal at $6.69 and oil at $3.69 per bushel, a board crush of $0.94. Oil was 36% of the product value. At USDA’s 2026/27 forecasts, oil is 53%.

The driver is biofuel demand. USDA ERS reports that soybean oil biofuel use for 2025/26 was raised to 14.7 billion pounds, and that EIA data showed 39% of US feedstock-based biofuel production used soybean oil in May 2026. The crush has become partly an energy spread: a processor’s margin now depends on renewable diesel policy and feedstock competition, not only on livestock feed demand for meal.

Real margins versus the board

USDA ERS also reports what processors actually extract and earn:

Extraction rate. In October 2024, US processors produced 11.91 pounds of oil per bushel, and ERS raised its 2024/25 forecast to 11.87 pounds. At 11.87 pounds and 70 cents, oil is worth $8.31 per bushel, not $7.70, which lifts the crush from $3.12 to $3.73.
Cash margins. ERS put the Central Illinois cash crush margin at $3.20 per bushel in October 2024, and its 2026 reports cite “favorable crush margins” as a reason for record crush forecasts.
Capacity. ERS estimated annual US crush capacity at 2.8 to 3.1 billion bushels (December 2024), and in August 2026 raised its 2026/27 crush forecast to a record 2.78 billion bushels.

As with refining, the board crush is a benchmark. The processor’s margin also depends on local basis for beans and products, energy and labour costs, hull and meal protein specifications, and freight.

Crush options

CME also lists options on the board crush. Per its guide, each option covers 50,000 bushels, strikes are in 2-cent increments, and the tick is one-eighth of a cent per bushel ($62.50 per contract). Exercising a call delivers a long crush (long 11 meal, long 9 oil, short 10 soybeans) at prices derived from the strike. Buyers pay a premium and do not post futures margin, which lets a crusher buy a floor under its margin, as a put.

What can go wrong

The spread can go negative. The gasoline crack averaged minus $5.44 in November 2008. Low margins can persist longer than a hedger’s patience.
Basis risk. A refinery’s crude and products are not WTI Cushing and New York Harbor. A crusher’s beans are not CBOT delivery-grade beans at Chicago prices. The hedge covers the benchmark, not the plant.
Yield mismatch. The 3-2-1 over-weights gasoline and distillate compared with EIA’s reported yields; the 10:11:9 under-hedges oil. Both errors move with prices.
Leg risk and margin calls. Each leg is margined and marked separately. A spread credit lowers the initial requirement, but daily losses on one leg are paid in cash even if the other leg gains.
Seasonal specification changes. The EIA notes that in summer the underlying of the RBOB front-month contract must meet a lower Reid vapor pressure specification, a more expensive gasoline: front-month prices step up from March 1 and the specification reverts on September 1. A crack built across those boundaries mixes winter and summer grades, which shifts its value without any change in refining economics.
Policy shocks. Biofuel mandates and fuel standards can reprice the oil share of the crush or the value of gasoline blending components quickly.

The short version

The 3-2-1 crack and the board crush turn processing margins into tradable futures positions. At our illustrative prices, the 3-2-1 was $21.60 per barrel, $64,800 per 3-2-1 set; at USDA’s 2026/27 forecasts, the board crush was $3.12 per bushel, or $149,000 per 10:11:9 set after the oil under-hedge. Both are approximations: real refineries yield less gasoline and diesel than 3-2-1 assumes, and real crushers extract more oil than 11 pounds. Official data shows how far these margins move, from a negative gasoline crack in 2008 to a record spot 3-2-1 in September 2026, and how biofuel demand pushed soybean oil past half of crush value. Exchanges give large margin credits for the spreads, but the legs still carry daily cash risk. This is a description of how the spreads work, not a trading recommendation.