67,994 Net Contracts Sit Inside a Gross Book Eight Times as Large

The Commitments of Traders report released on Friday 28 August 2026 shows non-commercial traders in E-mini S&P 500 futures net short 67,994 contracts. That single figure is the one that travels. It is also a difference between two columns that are each several times larger than it, measured three trading days before anyone outside the reporting system could read it.

Neither point makes the number useless. Both change how much weight it can carry. A net position is not a measurement of conviction. It is a residual left after one large column is subtracted from another, and a residual of that shape can move a long way without anyone changing their mind about direction.

What follows works through the arithmetic on one dated report using only the columns the CFTC publishes. Every figure below can be recomputed from the same two public sources, and the checks that confirm the columns are internally consistent take about ten minutes.

The Columns, and What They Add Up To

The file used here is the Legacy futures-only report for E-mini S&P 500 futures on the Chicago Mercantile Exchange, report date 25 August 2026. Its reported columns are these.

• Non-commercial: long 241,495, short 309,489, spreading 40,332
• Commercial: long 1,493,154, short 1,557,219
• Nonreportable: long 270,688, short 138,629
• Total open interest: 2,045,669

Before reading anything into those figures, two identities are worth checking, because a report that fails them has been transcribed wrongly somewhere. Reportable long positions are 241,495 + 40,332 + 1,493,154 = 1,774,981; adding nonreportable longs of 270,688 gives 2,045,669, the reported open interest. On the other side, 309,489 + 40,332 + 1,557,219 = 1,907,040, and adding 138,629 gives the same total. The CFTC states that the nonreportable line is not collected directly but is "derived by subtracting total long and short 'Reportable Positions' from the total open interest," which is why the columns close exactly rather than approximately.

The second identity is the more useful one. The three net positions are −67,994 for non-commercials, −64,065 for commercials, and +132,059 for the nonreportable residual. They sum to zero, as they must: a futures contract creates a long and a short at the same instant, so a category described as net short is being described relative to some other category that is net long by the same amount. On this report the category on the other side of both large groups is the one made up of positions too small to report.

Open interest itself is worth reading alongside the categories rather than after them. Total open interest fell from 2,072,358 on 18 August to 2,045,669 a week later, a decline of 26,689 contracts. Over the same week the non-commercial net moved by 57,434 contracts. The swing in the net was more than twice the change in the size of the whole market, which means the bulk of it came from contracts changing hands or changing category rather than from positions being opened. That distinction does not appear anywhere in the report; it comes from putting two of its own lines next to each other.

A Difference Between Two Large Numbers

The non-commercial category holds 241,495 long contracts and 309,489 short contracts. Its gross directional book is the sum of those, 550,984 contracts, which is 26.9% of total open interest. The net that gets quoted is 12.34% of that gross book — the gross is 8.1 times the size of the net drawn from it.

That ratio sets the sensitivity of the headline number. If 33,997 contracts moved from the short side to the long side and nothing else changed, the net would be zero. Those 33,997 contracts are 6.2% of the category's gross directional positions. A category can therefore rearrange a small fraction of its own book and produce a headline that reads as a complete reversal of stance, without a single contract of open interest being created or closed.

Spreading sits outside both columns above. The CFTC defines it for this report as a measure of "the extent to which each non-commercial trader holds equal long and short futures positions." At 40,332 contracts it is 14.3% of the non-commercial long side once spreading is included — a meaningful block of activity that is, by construction, directionally neutral and therefore absent from the net.

A low ratio of net to gross is not itself a defect in the data, and it is not evidence that the category is doing anything unusual. Two-sided books are what a category of many firms with separate mandates should produce. The consequence is narrower than that: it means the published net is a sensitive statistic, and sensitivity is a property to size before the number is used rather than after.

The Same Report, Eight Weeks Running

One reading in isolation says little about how much a reading of this kind normally moves. The eight most recent reports give the net as follows.

• 7 July −42,891 · 14 July −38,938 · 21 July −16,784 · 28 July −17,196
• 4 August −27,258 · 11 August +11,280 · 18 August −10,560 · 25 August −67,994

Seven week-over-week changes connect those eight readings: +3,953, +22,154, −412, −10,062, +38,538, −21,840, and −57,434. Ignoring direction, they average 22,056 contracts. The sign of the net changed twice inside eight weeks, and the largest single move, into the 25 August report, was more than twice the average.

Set that average against the current level and the scale problem becomes concrete. An average week moves this net by 22,056 contracts, which is 32.4% of the 67,994 the latest report shows. A statistic whose typical weekly change is roughly a third of its own level is not a slow-moving one, and treating a single reading as a description of settled positioning tends to overstate what it establishes.

Three Days You Cannot See

The CFTC's release schedule states that the reports are "released at 3:30 p.m. Eastern time," that they are "usually released on Friday," and that the release "usually includes data from the previous Tuesday." The schedule lists 7, 14, 21 and 28 August 2026 as release dates, and 4, 11, 18 and 25 September for the month following.

The snapshot is therefore taken at a Tuesday close and published at 3:30 p.m. on the following Friday. Wednesday, Thursday and most of Friday fall between the two — three of the five trading days that separate one snapshot from the next.

If a week's change accrued evenly across its five trading days, three days would carry 60% of it. At the average weekly change of 22,056 contracts that comes to about 13,234 contracts, or 19.5% of the 67,994 the report shows, moving inside the window before publication. Read the report at the open on the following Monday rather than at 3:30 p.m. Friday and the window is a full three sessions wide, with only Monday and Tuesday left before the next snapshot is taken.

The lag is a published feature of the report rather than a flaw in it. Reportable positions are filed daily by the firms that carry them, position by position, and the published report follows on a fixed schedule; three days is the cost of that process. What the lag rules out is one common use: reading a fresh print as a description of where a category stands at the moment of reading. It describes where the category stood at a specific past close, and the report says so on its face.

Two qualifications belong with that estimate rather than after it. Even accrual is an assumption, not a measured property; positioning tends to concentrate around scheduled events, so the true figure for any particular week may be far larger or near zero. And counting Friday as a full day before a 3:30 p.m. release overstates the window slightly, which makes 13,234 an upper bound within that assumption rather than a central estimate.

There is a practical consequence in how the series is read. Comparing this week's level against a memory of last week's headline compounds two lagged observations, while comparing two adjacent reports at least holds the lag constant: both were measured at a Tuesday close, one week apart. The change between two reports is a cleaner object than either level, because whatever staleness affects one affects the other in the same way. That is also why the week-over-week series above, rather than the level, was used to size the unobserved window.

The Category That Is Defined by Absence

The largest net on this report belongs to the nonreportable line, at +132,059 contracts. Nonreportable longs are 13.2% of open interest and nonreportable shorts 6.8%. This category is not a group of traders with a shared characteristic; it is everything left over once positions above the reporting level are removed.

Those levels are set in 17 CFR 15.03, which introduces its table with the phrase "The quantities for the purpose of reports filed under parts 17 and 18 of this chapter are as follows." Among broad-based security indexes the table lists 1,000 contracts for the S&P 500 Stock Price Index, 300 for the Municipal Bond Index, and 200 for other broad-based securities indexes. The E-mini contract has no line of its own, and that absence is the answer rather than a gap. When the Commission raised the E-mini level from 300 to 1,000 to match the full-size contract, it stated that "the reporting levels for the S&P 500 Stock Price Index contract and the E-Mini S&P 500 Stock Price Index contract will be the same," and that it was "deleting the separate reference to the E-Mini S&P 500 Stock Price Index in Rule 15.03" (69 FR 76392, 21 December 2004, effective 20 January 2005). The 200-contract line is a residual for indexes the table does not name; the S&P 500 Stock Price Index is named. The line that divides reportable from nonreportable in the columns above therefore sits at 1,000 contracts.

The classification of the two reportable categories carries its own limit. The CFTC states that "a single trading entity cannot be classified as both a commercial and non-commercial trader in the same commodity." An entity whose activity in a contract is partly hedging and partly not is placed wholly in one bucket. The categories are therefore administrative groupings, not descriptions of intent, and reading "non-commercial" as a synonym for a directional speculator attributes more to the label than it carries.

What Would Invalidate This

The ratio at the centre of this piece is specific to one contract in one week. Contracts where a category's gross book is close in size to its net produce a far more stable headline, because there is less offsetting activity for the difference to be drawn from. The 8.1 multiple should be recomputed per contract rather than carried across; it is one division on two published columns.

The eight-report window is short. Seven changes are not enough to characterise a distribution, and this particular stretch contains a sign change and one move of 57,434 contracts, which pulls the average up. A longer window would give a different figure, and the 22,056 used above should be treated as an order of magnitude rather than a constant.

The choice of file matters as well. Futures-only and futures-and-options-combined reports cover the same market and give different numbers, and the disaggregated and financial-futures reports split the same open interest into different categories. A net quoted without naming the file it came from cannot be checked.

Finally, none of the arithmetic here says anything about where the contract trades next. It bounds how much a positioning reading establishes, which is a separate question from direction and is not a substitute for one.

Concrete Framework

A checklist for reading a positioning figure before acting on it:

1. Name the file. Record which report the figure comes from — Legacy or disaggregated, futures-only or combined — and the report date, not the release date.
2. Check the identities. Reportable plus nonreportable should equal open interest on both sides, and the three category nets should sum to zero. If either fails, the transcription is wrong.
3. Divide the net by the gross. Add the category's long and short columns and compute the net as a percentage of that sum. At the 12.34% computed above, the headline is a small residual of a large book, and the lower that ratio runs, the less rearrangement it takes to move the headline a long way.
4. Compute the flip distance. Half the net is the number of contracts that must switch sides to zero it. Express that as a share of gross to see how small a rearrangement produces a reversal in the headline.
5. Measure the normal weekly change. Pull the last eight to twelve reports, take week-over-week changes, and average their absolute values. Compare that average to the current level.
6. Age the number. Count the sessions between the Tuesday snapshot and the moment of use. Multiply the average weekly change by the fraction of the week that has elapsed to size what is unobserved.
7. Separate the reading from the decision. Write down what the figure would have to show for a plan to change. If no threshold can be stated in advance, the figure is being used to confirm a view rather than to test one.

Positioning data describes what was held at a past instant by categories that are defined administratively. It is a constraint on interpretation more than an input to timing, and the arithmetic above is worth redoing on each contract rather than carried over as a rule of thumb.

This article is for general information about published market data and market structure. It is not investment advice, not a recommendation to buy or sell any contract or security, and it makes no claim about future prices.