Three numbers, and what each is worth
Funding, open interest and the liquidations of shorts are the whole public picture, and they are far more legible during a squeeze than before one. What you would need beforehand, how many shorts there are and at what price each one breaks, is published by nobody.
This page is about reading those three: where each comes from, at what grain it arrives, and what it leaves out. Why forced buying lifts a price at all is the mechanism, and it has its own page.
Where the three come from
| Number | Source, and the grain it arrives at |
|---|---|
| Funding, hourly and over a year | Hyperliquid's public API. The rate is recomputed every hour from a premium sampled every 5 seconds and averaged over that hour. On the funding page at the snapshot, and as a seven-day average on each market page. |
| Open interest, in dollars at the mark | Hyperliquid's public API gives the present only. Hyperkeel records it every hour, and each market page draws it since that record began. |
| Liquidations, by market and by side | Hyperliquid publishes no feed: on the public tape a liquidation looks like any other trade. Hyperkeel reads the fills of accounts that placed a large taker order, so every total is a floor. |
What the data shows before
Before a squeeze the three describe a market that could be squeezed. They do not single out one that will be, and no combination of them does.
- Funding below zero, and staying there. One negative hour is a snapshot; a market page's seven-day average says whether the discount held. By the documented formula the 8-hour rate goes negative only once the average premium index falls below -0.05%, so a negative print is already a real discount.
- Open interest building. Contracts were opened; the figure never names the side. In dollars at the mark it also rises with the price, so read the change over a short window.
- No count of shorts, at any grain. No public figure says how much of the open interest is short, nor the prices at which those positions would be liquidated. Anything offering such a figure has estimated it.
What the data shows during
While it is happening the picture is far less ambiguous. It also arrives as it happens rather than in advance, which is the whole difficulty.
- Liquidations of shorts clustering in one market. On the liquidations page the week is drawn every six hours, longs above the line and shorts below, with the largest listed by market, side, notional, price and time.
- Open interest falling while the price rises. Contracts closed into a rising price. That is also what ordinary profit-taking by longs looks like, and the pair cannot tell them apart.
- Funding turning positive. The discount that paid shorts becomes a premium charged to longs, and the switch can happen within one hour.
None of the three is timely enough to be a trade, and that is not a shortcoming of this site: a liquidation is already past tense by the time it appears in a fill, and an hourly record places a whole move inside one point.
What it cannot show
- Who is positioned how. No public figure splits open interest by side, by account or by entry price. Every contract is one long and one short.
- Where the forced buying would start. A liquidation price depends on the account's equity, its margin and whether the position is cross or isolated. Those are per account and not public.
- A total of liquidations. Hyperliquid publishes no feed. Hyperkeel counts liquidations above a threshold, found in the accounts it examined, so slices under that threshold are not counted and a total is always a floor rather than a sum.
- Intent. A large buy may be a liquidation, a stop, a hedge or a purchase someone had planned for weeks. In the data it is the same print.
Every figure here measures something that already happened, published at the grain of an hour or found after the fact. None of it is a signal, and nothing here is advice to take or avoid a position. Perpetuals are leveraged on both sides of a squeeze, and the whole deposit can be lost.