Two Minutes for Crude, Thirty Seconds for the Index: What a Futures Close Measures

A daily bar carries four numbers, and three of them are honest about what they are. The open, the high and the low each point at a price somebody paid. The fourth number carries almost all of the weight in systematic work — moving averages, breakout confirmation, trailing stops held overnight, every rule written as close above X — and on a futures chart it is frequently not a trade at all.

What sits in that field is a settlement price: a statistic the exchange computes from a defined window using a published procedure, then files with its regulator. The window differs by product. The procedure differs by product. For some contract months the number is derived rather than observed. None of this is hidden, and none of it requires a data vendor to explain it. It is written down in the exchange's own settlement documents and in the Commodity Futures Trading Commission's recordkeeping rules, and the whole of it can be checked in an afternoon.

This piece runs that check on two contracts — E-mini S&P 500 futures and NYMEX WTI crude oil futures — and pulls out the parts that change how the resulting series should be used.

The Rulebook Names the Gap Before the Exchange Does

Start with the regulation rather than the exchange, because the regulation admits the distinction in the plainest language available. Designated contract markets in the United States operate under core principles set out in Part 38 of Title 17 of the Code of Federal Regulations. Core Principle 8, at section 38.450, is one sentence: "The board of trade shall make public daily information on settlement prices, volume, open interest, and opening and closing ranges for actively traded contracts on the contract market."

Two things in that list matter for a chart. Settlement prices and closing ranges are named as separate items, and the closing item is a range rather than a point. A daily bar shows one close; the exchange is required to publish a range.

The recordkeeping rule is more direct still. 17 CFR 16.01(b)(2) tells each reporting market what to record for the opening and closing periods of trading. The first item is the opening and closing prices. The second item is "The price that is used for settlement purposes, if different from the closing price". That conditional clause is the whole story. The drafters expected the two numbers to diverge often enough that the rule needed a separate line for the case.

The same section anticipates two further cases that a chart cannot display. Where there were no transactions, bids or offers during the period, the market may record "Nominal opening or nominal closing prices that the reporting market reasonably determines to accurately reflect market conditions, clearly indicating that such prices are nominal." And where human judgment enters, the market must record "The method used by the reporting market in determining nominal prices and settlement prices", with the instruction that "Discretionary authority must be noted explicitly in each case in which it is applied (for example, by use of an asterisk or footnote)."

An asterisk beside a settlement price is therefore not typography. It is a regulatory flag stating that a person made a call on that print. Almost no charting package carries the flag forward, in the same way that a reported order-size bucket loses its definition once it is repackaged as a single number. The rule also fixes when the number arrives: settlement prices are filed "not later than 7:00 a.m. on the business day following the day to which the information pertains".

Equity Index Futures: Thirty Seconds on Globex

CME's client-systems documentation for E-Mini Standard and Poor's 500 futures sets out the normal daily procedure. Trades executed on CME Globex "between 14:59:30 and 15:00:00 CT, the settlement period" are volume-weighted, and the result is "rounded to the nearest .25 index point". Central Time runs one hour behind Eastern, so the instant that closes the equity index bar is 16:00:00 ET.

Thirty seconds is a short sample, and it is a sample of one venue. That is a design choice with consequences a chart user should hold onto. A large print at 15:00:04 CT, four seconds after the window shuts, moves the last trade and does not move the close. The two fields will disagree, and neither is broken.

Deferred months are usually not sampled at all. The same page derives settlements for the remaining months from a carry calculation built out of the index price, days to expiration divided by 365, and an interest rate. A back-month daily close produced that way is an arithmetic output. When the inputs move, the printed close moves with them, whether or not that contract traded during the session.

Then there is the version problem, which is the part most worth internalising. A CME rule filing hosted in the CFTC's filings directory describes this contract family with a settlement period running "15:14:30 – 15:15:00 Central Time". CME's own maintained page, marked as updated on August 20, gives 14:59:30 to 15:00:00 CT. Both documents are reachable right now, and they describe different windows. The operative rule is the one the exchange currently maintains; the working habit that follows is to record which document you read and what date it carried, because a settlement window is a parameter rather than a law of nature.

Daily settlement and final settlement are different objects

CME's equity index settlement material states the split plainly: "Daily settlement refers to the contract's settlement price on a daily basis while final settlement represents the final value of the contract at expiration." For the S&P 500 contract that final value is a Special Opening Quotation, which CME describes this way: "The SOQ is determined by the index provider and is calculated using the actual opening prices for each of the underlying constituent stocks."

An expiring contract's terminal value is therefore assembled from openings, not from any closing window, and it is assembled from cash constituents rather than from futures trades. A backtest that rolls on expiration day and marks the roll at the final settlement is marking against a number produced by a different mechanism from every other row in the series.

Crude Oil: Two Minutes, and Four Months That Are Inferred

The published NYMEX energy settlement procedure runs on a different clock. "The first six contract months in NYMEX WTI Crude Oil futures (CL), Natural Gas futures (NG), Heating Oil futures (HO), and RBOB Gasoline futures (RB) are settled by CME Group staff" using trading activity on CME Globex "between 14:28:00 and 14:30:00 Eastern Time (ET)". That is a two-minute window, four times the equity index sample, and it is stated directly in Eastern Time rather than Central.

Only the front month is settled from its own outright trades: it settles to the volume-weighted average price of the outright, "rounded to the nearest tradable tick". Everything behind it is inferred from spreads, and each rung of the ladder carries its own minimum volume.

Contract month	How the settlement price is formed	Minimum volume, crude oil
Front month	VWAP of outright trades in the window	Not applicable
Second month	Price implied from the front-to-second spread	200 contracts
Months three and four	Spread-implied, blended across two spreads	100 contracts
Months five and six	Spread-implied, blended across two spreads	One contract

Where both a one-month and a two-month spread are available, the document gives the blend: "an 85% weighting factor is applied to the price implied from the one-month spread, and a 15% weighting factor is applied to the price implied from the two-month spread".

Those thresholds belong to the product, not to the exchange. The same document sets the second-month minimum at 100 contracts for natural gas and 50 contracts for heating oil and RBOB gasoline. Carrying crude oil's 200 across to another energy contract would misstate the rule, and anyone building a multi-leg energy calculation is reading three of these documents at once, each with its own numbers.

Now place the two clocks side by side. Crude oil's settlement instant is 14:30:00 ET. The equity index settlement instant is 15:00:00 CT, which is 16:00:00 ET. The two settlement instants are 90 minutes apart. Any statistic that pairs a crude oil daily close with an equity index daily close — a correlation, a ratio, a spread, a regression residual — is pairing prices struck an hour and a half apart, on every row, for as long as the series runs. That is not an error in the data. It is a property of the data, and it is knowable before the first line of code is written.

Five Kinds of Number Can Occupy One Close Field

Collecting what the documents say, the value a vendor hands over as the close of a futures bar can be any of the following, and a chart draws all of them identically.

• A volume-weighted average of outright trades inside the settlement window. This is the normal case for a front month.
• A price implied from a spread quote, anchored on the month in front of it. This is the normal case for deferred energy months.
• A number produced by a carry calculation from an index level, a day count and an interest rate, with no trade involved.
• A nominal price, recorded where there were no transactions, bids or offers, and required to be clearly indicated as nominal.
• A determination made with discretion, required to be marked with an asterisk or footnote at the point where it was applied.

A series that mixes these is not wrong. It is heterogeneous, and treating a heterogeneous series as a homogeneous one is where the damage happens. The failure resembles applying a single threshold to two series whose units differ: the arithmetic runs, the backtest reports a number, and the number describes a quantity that was not uniform across the sample.

There is a second, quieter consequence. The settlement price is not only a chart value; it is the accounting value. CME's own settlement price fact sheet states that "Settlement prices are used to mark traders' positions to market on a daily basis, determining whether their positions had gains or losses on any particular day." The number that decides an overnight variation margin figure and the number a moving average consumes are the same number. The purpose stated in the source document is the first one; the second is a reuse.

What Would Invalidate This

Several conditions make the preceding analysis inapplicable, and each of them is checkable rather than a matter of opinion.

• Your vendor does not ship settlement prices. Some feeds populate the close field with the last traded price of the session instead. If that is the case, none of the window arithmetic above describes your series, and the correct move is to read the vendor's field definition before reading anything else.
• You are flat at the bell every day. For a trader flat at the bell, the settlement price is an accounting artefact of somebody else's margin call. The last trade may be the more faithful input, and the ninety-minute gap between products stops mattering.
• The product is not one of these two. Everything above is sourced from the E-mini S&P 500 procedure and the NYMEX energy procedure. Other CME products, and every other exchange, publish their own documents with their own windows and thresholds. Nothing here transfers by assumption.
• The document you are reading has no printed version date. The NYMEX energy settlement procedure carries no version date on its face, which is precisely why the conflicting equity index windows in two live documents are worth taking as a warning rather than as a curiosity.
• The chart is an equity chart. Cash equities reach their official close through a different mechanism entirely. None of the futures settlement reasoning above applies to a stock series, and borrowing it would be the same category error this piece is arguing against.
• Your signal does not read the close. A rule built on intraday bars, on volume, or on levels struck far from the settlement window may be entirely insulated from all of this. The test is whether any term in the rule dereferences a daily close.

Concrete Framework

Six steps, each producing a written artefact rather than an impression.

• Name the field. Write down the exact column name your data source uses and the definition it publishes for that column. If no definition is published, treat the column as unknown until one is found.
• Record the window per product, in one time zone. For each contract in the study, write the settlement window as the source states it, then a second line converting it to Eastern Time. Crude oil is 14:28:00 to 14:30:00 ET; the E-mini S&P 500 is 14:59:30 to 15:00:00 CT, which is 15:59:30 to 16:00:00 ET.
• Write the rounding rule next to the window. The equity index figure rounds to the nearest .25 index point; the crude oil figure rounds to the nearest tradable tick. A study that reports differences smaller than the rounding step is reporting the rounding step.
• Flag rows that were not outright trades. For deferred months, mark whether the settlement was spread-implied, carry-derived or nominal. If the source does not expose the flag, restrict the study to the front month and say so in the write-up.
• Keep a source register. Three columns: the URL, the date printed on the page, and the sentence you relied on. The equity index window changed between two documents that are both online; a register is what turns that from a surprise into a scheduled re-check.
• Re-check before any rule that references the close. Before a moving average, a breakout filter or a cross-product spread goes into production, confirm the two clocks still match the register. The re-check takes minutes; the alternative is discovering a ninety-minute misalignment inside a live result.

The general shape of the lesson is not new, but it is unusually well documented here. A close is not an observation of the market; it is a measurement, taken with a specified instrument, over a specified interval, by a specified party, with a specified rounding rule, and filed by a specified deadline. Every one of those specifications is public. Reading them takes an afternoon, and the alternative is a study whose most-used variable went undefined.

This is a note on reading published market data, not investment advice, and nothing above is a recommendation to buy or sell any security.