When Open Interest Rises on a Down Day, Four Readings Still Fit

Open interest went up is a complete sentence and an incomplete fact. It says that a certain number of contracts which did not exist at the previous close exist now. It does not say who wanted them, why they wanted them, or which way the holders think price is going. That gap is structural rather than a shortage of data. Every contract open interest counts has a long and a short attached to it, brought into existence at the same instant, at the same price, by two parties who disagreed.

The grid most traders learn — price up with rising open interest means new buying, price up with falling open interest means short covering, and so on — is usually taught as though it settled the question of who is doing what. It narrows the question. It does not settle it. What follows separates the part of the volume, open interest and price combination that is genuinely identified from the part that is inference wearing the clothes of observation, using the definitions the exchanges and the Commodity Futures Trading Commission publish rather than the ones that circulate in commentary.

Open interest counts contracts, not opinions

The CFTC states the constraint plainly in its explanatory notes to the Commitments of Traders report: “The aggregate of all long open interest is equal to the aggregate of all short open interest.” This is not an empirical regularity that happens to hold most of the time. It is an accounting identity. A futures contract is an obligation between two counterparties, and there is no arrangement under which more of them are long than short.

The consequence for reading a rising number is immediate. If open interest climbs by 10,000 contracts, 10,000 new long positions and 10,000 new short positions were created together. Nobody won that exchange at the moment it happened. The most that can be said is that a price was found at which one set of participants was willing to take on new exposure in one direction and another set was willing to take on new exposure in the other.

A single filled contract can move open interest three different ways depending on what the two sides were doing, and the tape shows the same thing in all of them.

Read the right-hand column of that table first. Volume rises by one in every row. Volume counts turnover — how much changed hands — and it counts closing trades exactly as it counts opening ones. Open interest counts the stock of positions that survived the close. The two are not two views of the same quantity, and a framework that treats a volume spike as confirmation of an open-interest move is quietly assuming a relationship that the definitions do not contain.

One further definitional wrinkle matters for physically delivered markets. The CFTC notes that open interest as reported to it “does not include open futures contracts against which notices of deliveries have been stopped.” Contracts heading into delivery leave the count. In a delivery month, part of any decline in open interest is the delivery process working normally rather than anyone changing a view.

The figure on the screen belongs to yesterday

There is a second problem, and it is the one most likely to be overlooked by someone building a rule around open interest: the number is not available while the session that produced it is running. Open interest is a cleared quantity. It exists only after the clearing house has matched and processed the day's trades.

CME Group is explicit about the two-stage publication. Its Daily Volume and Open Interest Report “is released at the end of each trading day and is a preliminary report,” and “CME Group releases official data in the Daily Bulletin the following morning.” The exchange adds the warning directly: “Preliminary reports may be different from the final report.” So there are two open interest figures for any trade day, they can disagree, and the one that stands is published the next morning.

Positioning data runs on a slower clock still. The CFTC describes the Commitments of Traders cycle this way: “Generally, the data in the COT reports is from Tuesday and released Friday. The CFTC receives the data from the reporting firms on Wednesday morning and then corrects and verifies the data for release by Friday afternoon.” The Commission publishes at 3:30 pm Eastern on Friday, using the immediately preceding Tuesday's positions.

Three full sessions trade between the snapshot and the publication. Whatever the report describes has had Wednesday, Thursday and most of Friday to change. That does not make the report useless — it is one of the few windows into position composition that exists at all — but it does rule out any reading that treats it as a description of the present.

What the three-way combination narrows

With those two constraints in place, the conventional grid can be stated honestly. Each cell contains a plausible story and at least three others that fit the same numbers.

Price	Open interest	Conventional label	What the data establish
Up	Up	New buying	Positions were created on net. Not who initiated them.
Up	Down	Short covering	Positions were closed on net. Not which side was forced.
Down	Up	New selling	Positions were created on net. Not the motive behind either leg.
Down	Down	Long liquidation	Positions were closed on net. Not whose.

Take the third row, since it is the one that most often gets read as a bearish confirmation. Price fell and open interest rose. At least four readings survive that observation:

• Speculative shorts initiating. The conventional story, and sometimes the correct one.
• Commercial hedging. A producer or inventory holder establishing a short hedge against physical stock is not expressing a directional view at all; the hedge is there so that a view is not required.
• Spread construction. A calendar or inter-commodity spread creates open interest on two legs at once. The trader is long one thing and short another and flat the outright direction.
• Mechanical hedging of an options book. Futures positions created to hedge option deltas are a function of the option position and the underlying price, not of an opinion about where price is headed next.

None of these four is exotic. In an actively hedged commodity market they run simultaneously, all day, and the published open interest change is their net.

Volume does contribute something the other two cannot, and it is worth isolating. Heavy volume against a nearly unchanged open interest is a statement about holding period. It means positions were opened and closed inside the session. That is a checkable fact about the horizon of the participants who were active, and it carries no directional content whatsoever — which is precisely why it is one of the more reliable things the combination reports.

A third of the book can be directionally neutral by construction

The Commitments of Traders report makes the composition problem concrete, because it breaks the same total into categories. Consider the Legacy Futures-Only report for WTI-Physical on the New York Mercantile Exchange, CFTC code 067651, for positions as of 18 August 2026.

Total open interest was 1,888,960 contracts, of 1,000 barrels each. That figure appears once, but it describes two entirely different books. On the long side, commercial traders held 894,821 contracts and non-commercial traders held 320,159 outright. On the short side, commercials held 1,047,908 and non-commercials held 198,069. The two sides balance to the contract, as the identity requires, and they look nothing alike.

The middle band deserves the most attention. The CFTC defines spreading as the extent to which a non-commercial trader holds equal long and short futures positions. That block was 598,872 contracts, 31.7 percent of open interest, and it is counted on both sides because it is the same positions viewed twice. Nearly a third of this market's open interest carried no outright direction as a matter of construction, not as a matter of interpretation.

The week-over-week change makes the same point from the other end. Open interest fell by 3,469 contracts from the previous Tuesday — a move of 0.18 percent, close enough to unchanged that most screens would not flag it. Underneath that flat headline, commercial short positions rose by 20,927 contracts and non-commercial short positions fell by 17,581. Roughly 1.1 percent of total open interest changed hands between categories while the aggregate barely moved. Anyone reading only the headline saw a quiet week.

One more caution applies to the combined reports. The CFTC states that in the futures-and-options-combined series, “option open interest and traders’ option positions are computed on a futures-equivalent basis using delta factors supplied by the exchanges.” The open interest figure in those tables is therefore not a count of contracts. It is a delta-weighted equivalent, and it will shift when deltas shift even if not one position changes. Comparing a combined-report figure with a futures-only figure compares two different quantities.

Expiration takes open interest to zero regardless

Every listed series has a terminal date, and on that date its open interest becomes zero. This is worth stating flatly because it means a declining open-interest curve into an expiry is the default condition, not a signal about conviction. The question is never whether front-series open interest is falling near expiry. It is whether the aggregate across all series fell, or whether the positions simply moved down the strip.

The clocks are not even synchronised between products written on the same underlying. Cboe specifies that trading in the SPX series on the S&P 500 index “will ordinarily cease on the business day (usually a Thursday) preceding the day on which the exercise-settlement value (i.e., the expiration date) is calculated,” at 5:00 pm ET, while trading in the SPXW series “will ordinarily cease on the day of expiration, 4:00 pm ET, and at 1:00 pm ET for any half day holiday.” Two contracts on one index, two different last trading conventions, two different days on which their open interest goes to zero.

Roll activity produces the same visual as capitulation and means the opposite. A position that leaves the front contract and reappears in the next one shows as a fall in one series and a rise in another, with the aggregate unchanged. Reading the front series alone during a roll window will show conviction draining out of a market where nothing has drained.

What Would Invalidate This

This frame is an argument about what can be inferred from an aggregate. Several conditions weaken it or remove it entirely.

• A homogeneous participant base. Where one category dominates a contract and the others are negligible, the aggregate change is much closer to identified, because there is only one plausible source for it. Check the trader counts in the COT before assuming heterogeneity: the WTI-Physical report above listed 283 traders in total across all categories.
• A single listed expiry. Markets without a strip have no roll confound. The decline-into-expiry problem does not arise.
• Higher-frequency position disclosure. If a venue publishes position data by account type more often than weekly, the lag argument shrinks in proportion.
• Liquidity questions rather than direction questions. Open interest is a good measure of how large the position base is, and therefore of how much size a market can absorb and how much could be forced to transact under stress. Nothing here argues against that use. The argument is narrowly about direction.
• Cash equities. There is no open interest in a share. Applying the frame outside listed derivatives is a category error in the first place.

Concrete Framework

A workable procedure for using open interest without over-reading it:

1. Fix which figure is being used. Preliminary end-of-day or official next-morning. Note that the exchange itself says these can differ, and never build a rule that needs an intraday value.
2. Aggregate the strip before drawing any conclusion. Sum open interest across listed expiries rather than reading the front series, particularly inside a roll window.
3. Check whether a delivery period is running. Contracts against which delivery notices have been stopped leave the reported count. Subtract that mechanism before calling a decline a loss of interest.
4. Confirm the series is comparable. Futures-only counts contracts. Futures-and-options-combined is delta-weighted. Do not chart one against the other.
5. Split the total by category when the COT covers the contract. Record the spreading share. If it is near a third of open interest, as it was in the example above, treat the headline number as a poor proxy for outright positioning.
6. Date the positioning data by its as-of date, not its release date. Tuesday's snapshot published Friday at 3:30 pm ET has three sessions of decay in it before it becomes public.
7. Write down the alternative readings before acting. For any price-and-open-interest cell, list the hedging, spreading and options-hedging explanations alongside the speculative one. If the intended action depends on which is true and there is no way to tell, the observation has not earned a position.
8. Keep the durable uses separate from the fragile ones. Position-base size, roll pressure and the amount of exposure that could be forced to transact are supported by the data. Direction is not.

The value of the volume, open interest and price combination is in what it rules out rather than what it points to. It can tell you that a move happened on position creation rather than position unwinding, and that distinction is real. It cannot tell you which side of the newly created pair was the eager one, because the number is defined in a way that makes both sides equal.