Open Interest Explained

What is open interest?

Open interest is the number of derivative contracts that have been opened and not yet closed. It is an accounting of outstanding obligations, measured as a level rather than a flow. Volume answers how much changed hands in a period; open interest answers how much risk is currently on the books.

The distinction is easiest to see in the extreme cases. A market can trade an enormous amount all day and finish with open interest exactly where it started, because every new position opened was matched by an old one closing — pure turnover, positions changing hands. Conversely, a quiet market can add a great deal of open interest on modest volume, because most of the trades that did occur were both parties opening. Those two days look similar on a volume histogram and describe completely different things.

One structural fact drives everything else: every contract has a long side and a short side, and open interest counts the pair once. There is no such thing as more longs than shorts. Statements of the form "open interest shows the longs are crowded" are not readings of open interest — they are readings of open interest combined with something else, and the something else is where the uncertainty lives.

How is open interest actually counted?

By what each party to a trade is doing, not by the trade itself. A single transaction has four possible effects on open interest depending on whether the buyer and the seller are each opening a new position or closing an existing one.

Two consequences follow, and both matter more than they first appear.

You cannot derive open interest from the trade tape. Nothing in a public trade print says whether either party was opening or closing. The venue knows because it maintains the positions; you know only because the venue publishes the number. Any tool showing open interest is relaying an exchange-reported figure, and it inherits that figure's cadence, units and revisions.

The published figure is a snapshot on the venue's own schedule. Open interest feeds publish on their own cadence, which is not the candle grid. The correct alignment is to take the latest sample whose timestamp is at or before a bar's close, and never to interpolate or back-fill from a later sample. Back-filling is straightforward lookahead — it hands a bar that closed at 10:03 an open interest value that did not exist until 10:05 — and it is one of the quieter ways a backtest becomes unreproducible.

The four open interest and price combinations

Pairing the change in open interest with the change in price produces the four-quadrant table that appears in every futures primer. It is genuinely useful, and it is an inference layer sitting on top of the accounting.

The genuinely definitional content of that table is thin and worth isolating: open interest up means net new positions, open interest down means net closing. That is all the number itself says. The words "longs" and "shorts" in the reading column are attributions of the new or departing risk to one side, and the attribution comes from price direction, not from the open interest figure, which is side-agnostic.

The most valuable rows are the two with falling open interest, because they identify moves that are structurally self-limiting. A rally driven by shorts closing consumes the very positions that produce it; when they are out, that source of buying ends without anyone changing their mind. A rally driven by new longs creates positions that must eventually be closed, which is a different kind of overhang. Same candle, different future obligations.

Why the four-way reading is a heuristic and not a law

Because each quadrant is a plausible dominant explanation for an aggregate net change, not the only one consistent with it. Open interest is a single net number over a period in which many participants did many things, and net numbers are compatible with many underlying compositions.

Concretely: price up with open interest up is consistent with new longs pushing price, and equally consistent with new shorts being opened into strength while other participants close — the net can rise either way. Hedging flow, basis trades and market-making inventory all add and remove open interest for reasons that have nothing to do with a directional view. On a venue where a large share of activity is delta-neutral, the directional reading of an open interest change is weaker than the table implies.

What tightens the inference is a third measurement. Aggressor-side delta — see CVD explained — says which side was paying to get filled during the period. Delta and open interest together are much harder to misread than either alone, which is the subject of the section below. Volume profile adds a fourth axis by saying where the business happened, and market structure a fifth by saying what price did with the levels it reached.

The honest formulation is therefore conditional. Rising open interest through a level means the move opened new risk; falling open interest through the same level means the move was substantially participants leaving. Both statements are supportable. "The longs are in control" is not, from this data.

What is different about open interest in crypto perpetuals?

Four things, and each of them changes how the number should be quoted. Perpetual futures have no expiry, so open interest never mechanically decays into a settlement date the way a dated contract does. There is no consolidated tape, so every figure belongs to one venue, and units differ between venues. The funding mechanism also means positioning has a continuous carrying cost that feeds back into whether positions are held.

Units are the trap. Open interest is variously published in contracts, in units of the base asset, or in quote-currency notional. A notional figure moves when price moves even if not a single position changed, because the same coin-denominated position is worth more. A notional open interest chart that rises alongside price is therefore partly an artefact of the denominator, and comparing it against a coin-denominated series is comparing two different quantities.

Venue fragmentation. Aggregate open interest across venues sums contracts with different specifications, margin regimes and reporting conventions. It is a useful shape and a bad absolute. If a price series and an open interest series come from different venues, the resulting overlay describes a shape rather than an audited count, and the honest thing to do is say so on the chart.

Positioning by price. A refinement worth knowing: instead of a single open interest line, allocate each bar's change in open interest to the price bin the bar traded in. That builds a profile of where open risk was created, which is a genuinely different question from where volume traded — a bar with heavy volume and flat open interest contributes nothing to it, while a bar with light volume and a large open interest jump contributes a lot. It shares the machinery of a volume profile and answers a different question with it.

Open interest together with order flow

The pairing that removes the most errors is delta against open interest, because it separates aggression from commitment. Delta says who was paying to get filled. Open interest says whether that aggression opened risk or removed it. The interesting cases are the ones where they disagree.

Strong buy-side delta with falling open interest is buying, but it is buying by participants closing shorts — the aggression is real and the conviction reading is not. Strong sell-side delta with falling open interest is the mirror: longs leaving, not new sellers arriving. Under rising open interest there is no divergence at all, because the aggression is opening fresh risk and the delta means what it appears to mean. A useful implementation therefore emits this divergence in one direction only, since the falling-open-interest case is the only one that contradicts the naive delta read.

This is the concrete version of the confluence idea: an engine that watches only delta will describe a short squeeze as buyers taking control right up until the moment it exhausts itself for no visible reason. Adding one independent axis makes that failure mode visible in advance rather than in hindsight, and it is the same logic that makes a liquidity sweep legible — a level taken on falling open interest was taken by participants leaving.

How does the Confluence Engine handle open interest?

The Confluence Show is a live, AI-run market analysis broadcast. Its analyst, NAIRO, reads raw trades, order books and positioning 24 hours a day through the Confluence Engine — 40+ analytical layers computed in-house — and positioning is one of the three raw inputs rather than an afterthought bolted onto a price model.

Three implementation rules follow from everything above. Open interest samples are aligned causally, taking the latest sample at or before a bar's close and never back-filling from a later one. A missing sample is reported as unknown rather than substituted with zero, because a zero would assert that no positions are open, which is a very different claim from not knowing. And the divergence layer between delta and open interest emits only where the two genuinely conflict, so the absence of an alert is itself informative.

NAIRO states in advance what would prove a positioning read wrong, and says on air when it was. This is educational market analysis, not financial advice, and it issues no signals or trade instructions of any kind. The covered assets are listed at /markets, the engine is described at /how-it-works, and the free 20-minute-delayed stream runs on YouTube, Twitch and Kick at /watch.