Rising open interest and falling price: what came next?
On 169 Hyperliquid coins, rising open interest with falling price was followed by a higher next day 51.44% of the time, vs 50.43% for all days.
On 169 Hyperliquid coins from June 23 to September 24, 2026, days with rising open interest and falling price were followed by a higher next day 51.44% of the time. The same measure for all days was 50.43%. The middle next-day move was +0.12% after those days, against +0.05% for all days.
Rising open interest with a falling price means more positions were opened than closed while the price fell. The textbook story says this means new shorts and a bearish market. In this sample, we did not see a bearish next day. These are past cases, not a forecast.
For what open interest is and how it differs from volume, see What is open interest in crypto?.
What are the four open interest and price combinations?
Open interest is the total size of trades that are still open. On a perp (a futures contract that never expires), it goes up when traders open new positions faster than others close, and down when positions close. Our open interest guide explains it in more detail.
Each day, a coin's open interest either rose or fell, and its price either rose or fell. That gives four combinations. Textbooks read them like this:
| Open interest | Price | Common reading |
|---|---|---|
| Up | Up | New buyers are pushing the price higher |
| Up | Down | New sellers (shorts) are pushing the price lower |
| Down | Up | Shorts are closing; the rise may be weak |
| Down | Down | Buyers are giving up; selling may be fading |
We wanted to know if those readings say anything about the next day.
How often did each combination happen?
Here is what we measured. We call one coin on one day a coin-day. Our 10,712 coin-days came from 169 coins.
| Combination | Coin-days | Share of coin-days | Next day higher | Middle next-day move |
|---|---|---|---|---|
| Open interest up, price up | 2,911 | 27.18% | 48.36% | −0.17% |
| Open interest up, price down | 2,268 | 21.17% | 51.44% | +0.12% |
| Open interest down, price up | 2,417 | 22.56% | 49.02% | −0.08% |
| Open interest down, price down | 3,052 | 28.49% | 52.88% | +0.20% |
| All coin-days | 10,712 | 100% | 50.43% | +0.05% |
The middle move (the median) is the middle result when you sort all the moves from lowest to highest. "Next day higher" counts days where the next UTC day closed above the day before. The next-day columns use the 9,382 coin-days that had a next day. The four rows do not sum to 100% because a tiny group where open interest barely changed (under 1% of coin-days) is left out. It is too small to read.
The pattern does not match the textbook. The "bullish" combination (up and up) had the lowest share of higher next days. The "bearish" combination (up and down) had a slightly higher share than all days.
Are those differences big enough to matter?
No. The differences from the 50.43% base are small: from 2.06 points below to 2.46 points above. The uncertainty on each share is about 3.5 to 4 points (for the up-and-down row, 3.75 points).
A standard check called a z-score measures how many units of uncertainty a result sits from the base. All four combinations were under 0.7 in size, and the up-and-down row was 0.27. Anything near 2 would be worth a second look. Nothing came close.
We also built the same table with only the busier coin-days (those with trades in at least 80% of the minutes). The up-and-down row came out at 51.69% higher, which is nearly the same. Weighting each coin equally rather than each coin-day gave 52.02%.
One more detail: the average next-day move for all coin-days was +0.56%, much higher than the middle move of +0.05%. A few very large moves pull the average up. That is why we show the middle.
What happened to liquidations on those days?
A liquidation is when an exchange closes a trade because the trader's money ran out. We also looked at the same day's liquidations in each combination, split into forced closures of longs (trades that gain when price rises) and shorts (trades that gain when price falls).
| Combination | Share of days with any liquidation | Average forced longs | Average forced shorts |
|---|---|---|---|
| Open interest up, price up | 47.8% | $17,518 | $57,003 |
| Open interest up, price down | 39.2% | $22,486 | $4,157 |
| Open interest down, price up | 40.3% | $9,105 | $27,815 |
| Open interest down, price down | 44.6% | $55,828 | $5,051 |
The middle liquidation total per coin-day was $0 in every group. Most coin-days had none. The averages are pulled up by a few big days.
The sides line up with price. Price-down days had more forced long closures, and price-up days had more forced short closures. That is expected, since a falling price pushes longs out. The biggest forced-long average, $55,828, was on days when open interest fell. Forced closures reduce open interest, so some of that fall may be the liquidations themselves.
The liquidations are from the same day as the move, not the next day. They help explain what the day looked like, not what followed. For which side gets forced out more often, see longs vs shorts liquidated.
What does this mean for the textbook reading?
The textbook reading is a story about who is on the other side of a move. Our numbers do not test the story directly. They test one claim that follows from it: that the next day tends to lean a certain way.
In this sample, it did not. The up-and-down combination, which the story calls bearish, was followed by slightly more higher days than the base. The gaps were small and well inside the noise.
Also, open interest can rise for two reasons that look the same in a total. A short seller can open a trade, and so can a buyer who is taking the other side of it. Every trade has two sides, so the total cannot say who is "driving" the move.
How we measured this
We used Hyperliquid perps from June 23 to September 24, 2026, which is 94 days. Open interest change is the dollar size opened minus the dollar size closed in each minute's trades, valued at the trade price, then added up over the UTC day. It is not the open interest level. Price move is the close-to-close return for the same UTC day. A day went into the "open interest flat" group if the change was under 1% of the coin's usual daily change over the previous 30 days; we left those out. The next-day result is the following day's close-to-close return. Of the 10,712 coin-days, 9,382 had a next day.
Limits: the window is short and sits in one market period. Coins move together, so days are not independent; we group the uncertainty by day. Bitcoin, Ethereum, Solana, HYPE, XRP and DOGE are not in the data, so this is the long tail of Hyperliquid coins, many thinly traded or new. Fees, spreads and funding are left out. These are past cases, not a forecast.
Try this in Comparand
With Comparand connected to your AI app, you can build the same table for a coin you care about:
- "Show the open interest change for one Hyperliquid coin each hour between two dates in September 2026." The open interest tool (hl_oi) returns the dollars opened, closed and net for each bar. You choose the bar size. It does not return the open interest level. It costs $0.005 per call on pay as you go and is unlimited on a plan under fair use.
- "Show that coin's liquidation totals per day for the same dates." The liquidation series tool (hl_liquidations_series) returns one coin's totals per complete UTC hour or day. It costs $0.005 per call on pay as you go.
- "Show that coin's price bars for the same days." The bar tool (hl_bars) gives the price side, so you can sign each day as up or down.
You then do the grouping and the next-day check yourself in your AI app. Every answer carries a prov block with the days and digests behind it. You can try 3 coins free, see the docs or check pricing.
For another way to read Hyperliquid positions, see our look at what happened after liquidation cascades. For why a small next-day lean should not be trusted on its own, see why past crypto moves are not a forecast.
The short version
- Rising open interest with falling price happened on 21.17% of 10,712 coin-days across 169 Hyperliquid coins (June 23 to September 24, 2026).
- The next day was higher 51.44% of the time after those days, against 50.43% for all days. The middle move was +0.12%.
- Every combination stayed within the noise of the base rate. The textbook "new shorts mean bearish" reading was not visible next day.
- Forced closures followed price: more forced longs on down days, more forced shorts on up days.
- This is a short window of thinly traded coins, without Bitcoin or Ethereum. Past cases, not a forecast.
Questions people ask
What does rising open interest with falling price mean?
It means more trades were opened than closed while the price fell. A common textbook reading is that new short trades are driving the drop and the move is bearish.
Is rising open interest with falling price bearish?
Not in our test. On 169 Hyperliquid coins from June 23 to September 24, 2026, the next day was up 51.44% of the time after such days. For all days it was up 50.43% of the time. The difference is far too small to tell from chance.
Which open interest and price combination had the best next day?
Open interest down with price down had the highest share of up days, at 52.88%. Open interest up with price up had the lowest, at 48.36%. Both are within normal noise for this sample.
Do these results apply to Bitcoin and Ethereum?
No. Our test left out Bitcoin, Ethereum, Solana, HYPE, XRP and DOGE because the data did not cover them. The results describe the long tail of Hyperliquid coins, many of them thinly traded.
Can a short test like this show the textbook reading is wrong?
No. It covers only 94 days in one market period, and days overlap across coins. It shows we found no next-day pattern in this sample, nothing more.