Skip to main content
In short: you want to know what it costs to hold a long on a perpetual (a futures contract with no end date), or what you could earn on the other side. That cost is funding, a payment that keeps the perpetual near the spot price. When it is positive, longs pay shorts. This tool gives daily funding for BTC or ETH back to 2021 from four exchanges side by side, and the share of days it went negative shows the risk of a trade built to collect it. Tool: get_funding_and_oi get_funding_and_oi returns the daily funding rate and open interest for BTC or ETH, back to 2021.

​
A positive rate is what longs pay

Funding is the payment that keeps a perpetual’s price near spot. Positive funding means longs pay shorts. So sustained positive funding is the running cost of holding the long side, and the figure to set against any yield a delta-neutral position claims to earn, because that yield usually is funding, collected on the short leg.
The share of negative days is the tail risk of the carry trade: a position built to collect funding loses money in exactly the periods when everybody wants to be short.

​
Four venues, never silently averaged

Binance, Bybit, Hyperliquid and Aster are each read separately and all four are reported. They agree closely (a correlation around 0.85 between Binance and Bybit on BTC), and where they do not, the disagreement is what to read. An average alone would hide exactly those days. Bybit has run about a point higher than Binance over the series. Quote whichever venue the position would trade on, not a blend. A rollup prints a mean alongside the venues rather than instead of them. It is equally weighted, because there is no per-venue open interest to weight it by; a venue holding none of the position would otherwise carry none of the average either. Only whole days count toward any average, and what counts as a whole day differs by venue: Hyperliquid settles hourly, Binance and Bybit every 8 hours. A day carrying one of three settlements is a third of a day’s cost.

​
Open interest is Binance’s book, not the market’s

Open interest and all three ratios come from Binance’s daily archive, which is the only one published here. Every positioning figure is therefore one venue’s position.
Say which venue when quoting one. “$8.4B on Binance" is a fact; "$8.4B of open interest” invites the reader to hear the whole market, which is not what was measured.
Funding is the only figure held from every venue, and the only one where they can be compared.

​
Rolling it up by period

group_by takes day, week or month and answers a different question from the daily series: whether open interest is building or unwinding, how one-sided positioning was last month against this one, and which market carries it. currency: "ALL" gives every market together, with each one’s own table underneath the combined one. Two things about the arithmetic affect how you read the output:
  • Open interest is a stock. A period’s figure is its mean, never a sum. Tuesday’s contracts are largely Monday’s, so a monthly total would be about thirty times anything a venue publishes. It does sum across assets, because two markets at one moment are two different positions.
  • An empty period is omitted, not zeroed. The archive is walked a day at a time, so a period can be partial; a partial one says so on its own row.

​
A missing series is not a zero

Open interest comes from a per-symbol daily archive that begins 2021-01-01, and some symbols are not in it. When one is absent the answer says so:
An open interest of zero says nobody holds a position at all, which is never true of a live market. If an agent reads a silent zero as a real value, it can conclude something dramatic about positioning from a hole in a data feed.

​
What comes back when it is there

Size in coins and in dollars, plus three ratios: top-trader long/short by position size and by account count, which differ when a few large accounts sit on one side, and taker buy/sell volume, which says who crossed the spread. They describe positioning that has already happened, and they are not trading signals.