How to Read Open Interest
It is not a direction indicator and it never was. It is a size indicator, it only speaks at one end of its range, and that one reading is worth having.
By Pavel Penev, MScFounder, TradeWize · 10+ years trading the marketsThe short answer
Open interest is the number of contracts currently open — bets that have been placed and not yet closed. It is a measure of how much money is standing on the table, and that is the whole of it. It does not say which way the table is leaning. What 2,160 days of bitcoin do say is that when open interest sits more than 15% above its own 30-day average, the following week runs a 17.1% high-to-low range against 12.4% normally, and 37% of those weeks take a drawdown past 10% against 16%. Everywhere else in its range it says nothing at all.
Open interest sits on every crypto dashboard next to the funding rate, and it gets read the same confident way. Open interest is at an all-time high, so the market is over-leveraged and due a flush. Open interest is falling on this rally, so there is no real money behind it. Both sentences are said constantly and neither is tested.
So we tested them. Daily open interest on the three biggest perpetual contracts — Bitcoin, Ether, Solana — from Binance's own public archive, joined to what price did over the following 7 days. 2,160 days of bitcoin, 1,703 of ether, 1,702 of solana. The famous readings are not in there. Something else is.
What it actually counts
Every futures contract is an agreement between two people. One of them is long, one of them is short, and the contract exists from the moment they are matched until one of them closes it. Open interest is the count of those agreements that are still alive right now.
That makes it a different animal from volume, and the difference is the single most useful thing to get straight. Volume counts trades. Open interest counts positions. If you buy a contract and sell it an hour later, you have added twice to today's volume and nothing at all to open interest — you opened one contract and closed it. If you buy and hold, you have added one to both.
Volume
How much changed hands today
- Resets to zero every session.
- A day-trader who opens and closes ten times adds ten times over.
- Tells you how busy the market was.
Open interest
How much is still on the table
- A running total that carries across days.
- The same day-trader ends the session having added nothing.
- Tells you how much money is committed and exposed.
Which is why it matters more in leveraged markets than anywhere else. A live perpetual contract is somebody's borrowed money sitting on somebody's margin. When there are a lot of them, there is a lot of borrowed money that can be forced to close in a hurry — and a forced sale does not wait for a good price. That is the mechanism behind everything below: not a crowd that has taken a view, but a crowd that can be made to act all at once.
First, get the unit right
Almost every dashboard shows open interest in dollars, and that is the wrong unit for the question you are asking. The dollar figure is the number of contracts multiplied by the price, so in a rising market it goes up when nothing has been opened at all. Read it in dollars and you will call an all-time high that is really just the price.
Bitcoin makes the point better than an argument can. Its open interest in dollars set its record on 6 October 2025 at $12.48bn. Its open interest in contracts set its record on 26 October 2022 — nearly three years earlier, with bitcoin at $20,758, in the pit of a bear market. There were 166,785 contracts open that day against 100,137 on the dollar record. The most crowded the book has ever been and the most valuable it has ever been are three years and a bear market apart.
Bitcoin open interest on Binance USDⓈ-M futures, the highest reading of each month since September 2020. The gold line is what your dashboard shows and it tracks the price. The cyan line is the position itself. Everything in this article is measured on the cyan one.
So compare it to itself, not to a number
A contract count means nothing on its own: 112,964 open contracts is the whole bitcoin book and would be an impossible number on a small altcoin perp. Every reading in this article is open interest divided by its own mean over the previous 30 days, so 1.00× is that contract's own recent normal and 1.20× means a fifth more is open than the market has been carrying. That is the form the number is worth looking at in, and it is two clicks on any charting platform.
Now the test everybody skips
Signal: the day's open interest divided by its own mean over the previous 30 days, so a reading is judged against that contract's recent normal rather than against a dollar level. Open interest is counted in CONTRACTS, not in dollars, so the number does not move just because price did. Outcome: the following seven days — the change in the close, the high-to-low range as a share of the starting price, and the deepest drawdown from it. Every day in the sample is an observation, so the windows overlap.
Sort those days into ten equal slices by the reading and look at what the following week did. If open interest is the dial it gets treated as, the slices should march.
Each bar is the mean high-to-low range of the following 7 days, by decile of the reading. Deciles 1 to 9 sit between 11.1% and 13.6%. The tenth runs 15.7%.
They do not march. Nine of the ten deciles are the same week: the range never leaves a 2.5%-wide corridor, and the share of weeks taking a drawdown past 10% sits between 12% and 20% in every one of them. Only the top decile separates, and it separates hard: 15.7% range and 28% of weeks going past a 10% drawdown.
Across the whole sample the rank correlation between the reading and the following week's range is +0.02, which is nothing. Both facts are true at once and the second one is why the first one is worth trusting. Open interest is not a dial you read continuously. It is a threshold that stays quiet until it is crossed.
What the top band actually says
Group the same days into five plain readings instead of ten slices and the shape survives. Four of the five bands are the ordinary week. The fifth is not.
| Open interest vs its own average | Share of days | Next week's range | Mean drawdown | Drawdown past 10% | Higher a week later |
|---|---|---|---|---|---|
| below 0.92× | 14% | 13.5% | −6.73% | 20% | 50% |
| 0.92× – 1.00× | 29% | 11.7% | −5.03% | 12% | 59% |
| 1.00× – 1.06× | 34% | 11.6% | −5.55% | 15% | 54% |
| 1.06× – 1.15× | 17% | 12.4% | −6.18% | 17% | 48% |
| above 1.15× | 5% | 17.1% | −9.23% | 37% | 44% |
BTCUSDT on Binance USDⓈ-M futures, 2020-10-01 to 2026-08-31, 2,160 overlapping observations. Every week in the sample: 12.4% range, −5.87% mean drawdown, 16% of weeks past a 10% drawdown, higher 53% of the time.
Read the last row against the sample. The range goes from 12.4% to 17.1%. The mean drawdown goes from −5.87% to −9.23%. And the number that will actually change how you size a trade: the share of weeks that dropped more than 10% below where they started goes from 16% to 37%. More than double.
Note what is NOT in that row. The last column barely moves — 44% against 53% — and the mean week is −0.29% against +0.98%. That is a market that has become harder to sit in, not one that has decided to fall.
Notice too that the emptying row is the second-widest. That is not a prediction, it is an echo: open interest collapses during violent moves, so a low reading is often the day after a crash and the volatility has not gone home yet. The shape is a lopsided U, and only the right arm reproduces.
Which is the next thing to check, because one contract proving a point is not a point.
| Contract | Since | Crowded days | Range: crowded vs all | Drawdown past 10%: crowded vs all | Mean week |
|---|---|---|---|---|---|
| Bitcoin | 2020-10-01 | 117 | 17.1% vs 12.4% | 37% vs 16% | −0.29% vs +0.98% |
| Ether | 2021-12-31 | 127 | 16.7% vs 14.5% | 35% vs 25% | −0.60% vs +0.30% |
| Solana | 2021-12-31 | 223 | 25.5% vs 20.4% | 44% vs 38% | +0.94% vs +0.69% |
Bitcoin's archive starts in September 2020; the other two start in December 2021, which is where the public metrics archive begins for them. The magnitude columns move the same way on all three. The last column does not: bitcoin's crowded week is −0.29%, ether's is −0.60% and solana's is +0.94%.
That last column is the honest end of this article. Two contracts out of three point down and the third points up, which is not a finding, it is a coin. The magnitude columns move together on all three and by similar amounts, which is. So the claim worth carrying is the narrow one: a crowded book makes the next week bigger in both directions, and has nothing to say about which one you get.
Learn it by doing
Reading about it is one thing — it clicks when you do it. Learn it hands-on with free, interactive lessons on TradeWize.
Try the free lesson →The four-box table in every textbook
There is a grid that appears in almost every introduction to futures. Open interest rising or falling, crossed with price rising or falling, four boxes, each with a verdict. Rising open interest with a rising price is the strong trend with fresh money behind it. A rally on falling open interest is short covering and not to be trusted. It is tidy, it is memorable, and it is on the wall of every trading course.
| The move | What it is called | What the textbook says it means |
|---|---|---|
| Open interest up, price up | New longs | The strongest signal in the grid — fresh money funding a real trend. |
| Open interest up, price down | New shorts | Fresh money betting on the downside. Treated as a strong bear trend. |
| Open interest down, price up | Short covering | A rally with no new money behind it. Treated as weak and unreliable. |
| Open interest down, price down | Longs closing out | A sell-off that is running out of sellers. Treated as a fading move. |
Run it. Take the three-day change in open interest and the three-day change in price, put each day in its box, and look at the following week. The grid should order itself the way the verdicts say.
| The move | Bitcoin: higher a week later | Ether: higher a week later | Solana: higher a week later |
|---|---|---|---|
| Open interest up, price up — new longs | 50% (n=666) | 46% (n=526) | 47% (n=450) |
| Open interest up, price down — new shorts | 55% (n=503) | 50% (n=397) | 51% (n=407) |
| Open interest down, price up — short covering | 55% (n=482) | 51% (n=346) | 48% (n=387) |
| Open interest down, price down — longs closing out | 54% (n=509) | 51% (n=434) | 46% (n=458) |
Share of observations where price was higher seven days later. For reference, the base rate on each contract is Bitcoin 53%, Ether 49%, Solana 48%. The spread between the best and worst box is a few points on every contract, and the ordering is different on each.
The grid does not hold. Nothing orders the same way twice, and the box it rates highest — open interest up, price up, the healthy well-funded trend — finished last on bitcoin, last on ether, third of four on solana. Its win rates were Bitcoin 50%, Ether 46%, Solana 47%, against base rates of 53%, 49%, 48% — the only box on the grid that was worse than doing nothing on all 3.
It is worth being clear about why, because the grid is not stupid. Its four descriptions are accurate: rising open interest with a rising price genuinely is new longs, and a rally on falling open interest genuinely is people buying back shorts. The mechanics are right. What is wrong is the leap from the mechanics to a verdict — from "this is what just happened" to "so this is what happens next". Knowing that fresh longs funded the last three days tells you what the position looks like. It does not tell you what the position is about to do.
Set today’s open interest against its own 30-day average and the panel says what the following week looked like across 2,160 days of BTCUSDT.
BTCUSDT on Binance USDⓈ-M futures, 2020-10-01 to 2026-08-31, read on 2026-09-09. Open interest is counted in contracts, not dollars, so the reading does not move just because price did. Forward windows overlap and ignore fees, so read the panels as a shape rather than as numbers to trade. Now drag the slider end to end: the top panel moves by about a third and the middle one nearly triples, while the bottom one wanders without an order. The same top band widens the following week on ether (16.7% against 14.5%) and on solana (25.5% against 20.4%) — and points a different way on each.
Today’s reading, for reference: 106,627 contracts against a 30-day average of 107,796 — 0.99×, which is below normal. 3 contracts were studied in all.
The week the other number could not see
Our article on the funding rate ran a table of the days either side of 10 October 2025 — the largest liquidation event crypto has had, about $19bn of positions force-closed inside 24 hours across 1.6 million accounts. The funding rate, the number everybody watches for exactly this, sat below its own base rate through the run-in and was falling as the day approached. It saw nothing, because it measures what positions are being paid for right now.
Here is the same week, measured with this number instead.
Bitcoin open interest on Binance USDⓈ-M futures, with the closing price above it. The build-up is visible for a week; the unwind takes one afternoon.
| Date | Contracts open | Value | vs its 30-day average | Close | Low |
|---|---|---|---|---|---|
| 3 Oct | 95,866 | $11.71bn | 1.08× | $122,183 | $119,200 |
| 4 Oct | 96,414 | $11.80bn | 1.08× | $122,343 | $121,432 |
| 5 Oct | 99,289 | $12.25bn | 1.11× | $123,427 | $122,088 |
| 6 Oct | 100,137 | $12.48bn | 1.11× | $124,628 | $123,018 |
| 7 Oct | 93,161 | $11.30bn | 1.03× | $121,286 | $120,516 |
| 8 Oct | 92,252 | $11.37bn | 1.02× | $123,238 | $121,005 |
| 9 Oct | 96,730 | $11.76bn | 1.07× | $121,579 | $119,573 |
| 10 Oct | 72,614 | $8.19bn | 0.80× | $112,715 | $101,516 |
| 11 Oct | 74,334 | $8.22bn | 0.83× | $110,579 | $109,501 |
| 12 Oct | 75,918 | $8.72bn | 0.85× | $114,894 | $109,509 |
| 13 Oct | 76,673 | $8.83bn | 0.86× | $115,112 | $113,600 |
| 14 Oct | 79,285 | $8.96bn | 0.90× | $112,984 | $109,802 |
Last reading of each day. The dollar record on 6 Oct is the all-time high of the whole 73-month series.
Open interest climbed for a week and set its all-time dollar record on 6 October 2025, four days out, at 1.11× its own average. It dipped, then rebuilt to 1.07× on the 9th. Then, on the 10th, 25% of the standing position closed in a single day — 27% measured from the record — while price traded from $121,579 down to $101,516.
One number was blind and the other was not, and that is the lesson
Funding measures the price being paid to hold a position right now — a flow, and it can be tiny while an enormous amount of borrowed money sits in the market. Open interest measures how much is sitting there — a stock, and it was at its record. Neither is a crystal ball, and open interest did not say the 10th. What it said, correctly and for a week beforehand, was that whatever happened next was going to be violent.
And be honest about the sample this sits in. There have been 117 crowded days on bitcoin in six years, belonging to 26 separate episodes, and October 2025 is the most spectacular of them. One vivid case is a story, not evidence — the evidence is the three-contract table further up, where the effect shows without needing a famous week in it.
What could be wrong with all this
A result this clean deserves the argument against it stated properly, so here it is.
The honest limits of this test
One venue. Binance is the largest perpetual venue but not the market; open interest on Bybit, OKX, Hyperliquid and the CME is not in here, and a reading that is extreme on one venue can be ordinary across the rest. Overlapping windows. Consecutive observations share most of their outcome, so the sample is far less independent than the day count suggests — the non-overlapping check below is the honest sample size. Two of the three contracts start in December 2021, because that is where Binance's public metrics archive starts for them. Only bitcoin has the full six years. In-sample and gross. No fees, no funding, no slippage, and no attempt to trade any of it.
The overlapping-windows point is the one that bites hardest, so it gets checked rather than mentioned. Cut the sample so no two observations share a day and 2,160 rolling weeks become 309 independent ones. The result survives — the top band still runs the widest range and the deepest drawdown of the five, 13.9% against 12.3% and −9.39% against −5.85% — but it survives on 17 weeks. That is enough to say the finding did not evaporate. It is not enough to quote to two decimal places, which is why this paragraph exists.
There is a second one worth naming. This is one venue, and open interest is fragmented across a dozen of them plus the CME. A reading that looks extreme on Binance can be ordinary once the rest of the market is added, and the aggregate figures the data providers sell exist precisely because of that. Nothing here should be read as "the market's open interest" — it is the largest single book's, which is the closest thing to it that anybody can check for free.
How to actually read it
- Read it in contracts, not dollars. The dollar figure is contracts × price, so in a bull market it makes records out of nothing. Bitcoin's contract record and its dollar record are nearly three years apart.
- Compare it to its own recent average, not to a level. 1.00× is that contract's normal. The only reading in this study that changes anything is above 1.15×, and on a different contract that is a completely different dollar number.
- Treat it as a size forecast, not a direction one. Crowded means the next week ran 17.1% instead of 12.4% and went past a 10% drawdown 37% of the time instead of 16%. It does not mean down.
- Act on it with your position size and your stop distance, which are the two things a size forecast is actually for. A stop that is comfortable in a normal week is inside the noise in a crowded one.
- Ignore the four-box grid. Rising open interest on a rising price is genuinely new longs — and that box was the only one of the four with worse odds than the contract’s own base rate, on every contract tested here.
- Watch the fall as closely as the rise. A drop of 25% in a day is not people changing their minds, it is people being closed out, and it tells you the leverage has already been flushed.
- Pair it with funding rather than choosing between them. Funding says which side is crowded and what it is paying; open interest says how much is standing there. The October 2025 week is the case for reading both.
As this was written, bitcoin's perpetual carried 106,627 open contracts, worth $8.36bn at $79,272, against a 30-day average of 107,796 — a reading of 0.99×. Quieter than usual. Which is where the number sits 63% of the time, and the reading worth getting comfortable with.
The one-line version
Open interest counts the positions that are still open, and it answers one question well: how much is on the table. Above 1.15× its own 30-day average, the next week on bitcoin ran 17.1% wide against 12.4% and went past a 10% drawdown 37% of the time against 16% — and that reproduces on ether and solana. Which way it went reproduces on nothing. Size for the first fact and stop trying to trade the second.
What is open interest?
Open interest is the number of futures or options contracts that have been opened and not yet closed. Every contract has a long and a short, and it counts as one until one of them exits. Unlike volume, which counts trades and resets each session, open interest is a running total of positions that still exist — so it measures how much money is committed to the market rather than how busy the market was.
What's the difference between open interest and volume?
Volume counts every trade in a session and resets to zero the next day; open interest counts positions still open and carries across days. A trader who buys and sells within an hour adds twice to volume and nothing to open interest. High volume with flat open interest means the day was churn; rising open interest means new positions were genuinely added.
Does high open interest mean a crash is coming?
Not in this data — it means a bigger week is coming, in whichever direction. When bitcoin's open interest sat more than 15% above its 30-day average, the following week's high-to-low range averaged 17.1% against 12.4% for the whole sample, and 37% of those weeks fell more than 10% below their starting point against 16%. But price was higher a week later 44% of the time against 53% — barely different. The size changes; the direction does not.
Should I read open interest in dollars or in contracts?
In contracts. The dollar figure is contracts multiplied by price, so it sets records in a bull market when nothing new has been opened. Bitcoin's open interest hit its dollar record on 6 October 2025 with bitcoin at $124,628, and its contract record on 26 October 2022 with bitcoin at $20,758 — nearly three years apart, and the second one was in a bear market.
Does rising open interest with a rising price confirm the trend?
That is what the standard four-quadrant table says, and it did not survive the test. On bitcoin that combination was followed by a higher price a week later 50% of the time against a base rate of 53%; on ether 46% against 49%; on solana 47% against 48%. It finished last on bitcoin, last on ether, third of four on solana. The description is accurate — it really is new longs — but the verdict attached to it is not supported.
What does falling open interest mean?
Positions are being closed, either voluntarily or by force. It is usually the aftermath of a fast move rather than a forecast of one: on 10 October 2025, 25% of bitcoin's standing position closed in a single day as prices fell. In this study the lowest band was the second-widest of the five for the following week, which is the echo of whatever emptied it, not a signal about what comes next.
What is a high open interest reading?
There is no absolute number, which is why the reading has to be relative. Divide open interest by its own average over the previous 30 days: 1.00× is normal for that contract, and the only band in this study that behaved differently was above 1.15×, which happened on 5% of days over six years.
How is open interest different from the funding rate?
Open interest is a stock and funding is a flow. Open interest says how much borrowed money is standing in the market; funding says what the crowded side is paying to keep it there right now. They can disagree completely — in the run-in to October 2025 funding sat below its own base rate while open interest was at an all-time high — which is the argument for reading both rather than picking one.
Open interest sizes the move. The chart shows you where it happens
Upload a screenshot or type a ticker and the chart analyzer reads the levels, the trend and the structure off the real candles — including where a crowded book's stops are likely sitting. It never tells you what to buy. It tells you what the chart is doing, so a decision about size is one you make with the picture in front of you. Free, no account, one read a week.