Methodology

Where every number comes from

Your browser asks Hyperliquid's public API (api.hyperliquid.xyz) directly, on your own connection. Nothing passes through our servers. Two feeds carry every market at once over one connection; everything else is requested when a page needs it.

• Hyperliquid limits how much one address may request per minute. Each browser tab spends at most about 600 units a minute through one queue, waits when Hyperliquid asks it to, and slows long requests (90-day histories, measured books) to stay inside the limit.

Price

The mark price Hyperliquid reports for the market. For spot, in the pair's quote token.

Change over 24 hours

How far the price moved over the last 24 hours, rolling, not since midnight.

• Checked 2026-10-04: BTC's prevDayPx (84,978) equalled its price 24 hours earlier, not at 00:00 UTC (84,716).

Volume over 24 hours

The value traded over the last 24 hours, rolling, in the market's settlement token: USDC for every perp exchange with live markets (2026-10-04); for spot, the pair's quote token.

• Checked 2026-10-04: BTC's 24-hour base volume (7,132 BTC) lay between its volume since 00:00 UTC (5,932) and over the last 25 hours (7,702), so the window rolls.

Open interest

The value of a perp's open contracts: Hyperliquid's open-interest figure, in contracts, times the mark price.

• We read the figure as one side of the market: every long has a short, and each pair is counted once. The evidence is the caps (see cap use): read this way no market is over its cap; counting both sides would put six over (2026-10-04).
• Other sites may count open interest differently and show other totals. This is our definition, not a judgement of theirs.

Funding rate

The rate longs pay shorts (or shorts pay longs, when negative) at the end of the current hour, as a share of the position's value. Per year is the hourly rate × 24 × 365, not compounded.

• Main exchange: interest 0.01% and clamp 0.05% per 8 hours. Builder exchanges set their own; where they have not, the clamp defaults to 0.03% (documented), and the interest rate (0.01%) and multiplier (1) are inferred from live rates, not documented. On 2026-10-04, 49 builder markets sat exactly at the base rate those defaults give.
• Hyperliquid's own predicted rate is not shown separately: on 2026-10-04 it equalled this figure on 154 of 178 main-exchange perps and differed on the rest only by when each was sampled.

Base rate, premium and the band

While the hour's average premium stays within the clamp of the interest rate (the band), funding is the base rate: the interest rate, times the multiplier on builder exchanges. Outside it, the premium sets the rate. "To band edge" is how far the latest premium sample sits inside (+) or outside (−) the band.

• One sample and the average can disagree: on 2026-10-04 they did for 50 of 330 perps.
• Builder exchanges measure the premium from the midpoint of the impact prices, without the main exchange's band around the oracle.

Mark against oracle

How far the mark price sits from the oracle price, in basis points (hundredths of a percent).

Impact spread

What buying and immediately selling a set size costs, before fees: the gap between the average prices to buy and to sell that size at once, as a share of their midpoint.

• The size differs: $20,000 on BTC and ETH and $6,000 on other main-exchange perps (Hyperliquid's contract specifications); builder exchanges do not publish theirs, so their figures may not compare exactly.
• Checked 2026-10-04 on 22 main-exchange perps against their live books: the impact prices matched the average fill for $6,000 to within about 1 bp, and were 3–20 bp from the price of the last level reached.

Funding against Binance and Bybit

Binance's and Bybit's next predicted funding for the same coin, put on an hourly basis, and the gap between Hyperliquid's rate and their average.

• An entry without an interval is left missing rather than assumed 8-hourly (20 Binance entries on 2026-10-04).

Open-interest caps and cap use

Each builder exchange caps every market's open interest in dollars. Cap use is open interest against that cap. The main exchange publishes no caps; it lists the markets at their cap.

• Hyperliquid's documentation defines the caps as notional ("position size times mark price"). That cap use counts one side of open interest is our reading of live data: no market was over 100% that way (highest 85%, 2026-10-04).
• perpDexs and perpDexLimits gave identical per-market caps on all 10 builder exchanges (2026-10-04). The main exchange's at-cap list also holds delisted markets, which are left out.
• "At or near cap" means 80% or more of a builder market's cap, on every page.

Exchange-wide cap

The cap a builder exchange sets on its total open interest, in dollars, and how much of it is used.

• perpDexLimits also returns a per-market size cap and a transfer limit, not shown: we cannot explain them from the documentation.

Open interest against volume

Open interest divided by the last 24 hours' volume: above 1×, more is held open than traded in a day. A rough sign of how long positions stay open.

Order-book depth

The value of orders resting within a band around the mid price, on one side of the book.

• Past the live levels the books are grouped into price buckets, priced outward (bids rounded down, asks up), so depth there reads low: shown as "at least". Where no book reaches the band, it says "not measured", never zero.
• Each figure is taken from the combination of books that reaches furthest; no single combination was best everywhere (2026-10-04).

Cost to trade

How far the average fill of a market order of a given size lands from the mid price, one way, in basis points.

• Bounds as for depth: "at most" past the live levels, "not measured" where the book is too thin.
• The impact spread covers a buy and a sell, so at the same size it is about the buy and sell costs added.

Spread

Best bid to best ask, as a share of the mid price, in basis points.

Funding history and what a long paid

The hourly rates applied in the past. "Paid by a long" is their sum over the period: the share of a long position's value paid in funding, if its value stayed the same (negative: received).

• fundingHistory returns the first 500 hours after the start, not the latest, so longer periods are read page by page.
• Hours are stamped a few milliseconds after the hour (4–90 ms, 2026-10-04) and are aligned by rounding.

Lending

Hyperliquid's built-in lending: per token, supplied and borrowed (in tokens, and in dollars at the reply's oracle price), the share lent out, and the yearly supply and borrow rates.

• Checked 2026-10-04: all five tokens' live rates matched the formula within 1e-11. Every token was below 80% lent out, so the steep part above it rests on the documentation alone.
• A token whose rates leave the formula is shown as reported, without the explanation.

Listing auction

The Dutch auction for the right to list one more perp, shared by all builder exchanges: the price, in HYPE, falls in a straight line for 31 hours to 500 HYPE, and the first to pay it gets the listing.

• Checked 2026-10-04: reported 752.52 HYPE, the rule gave 752.51. If the reported price leaves the line, it is shown as reported, without the explanation.

Leverage and margin

The most leverage a position may use, lower for larger positions where the market has tiers, and whether it may share margin with other positions.

• Tiers are by position value in USDC (Hyperliquid's margin tiers: BTC 40× up to 150M, 20× above). A market whose table is not listed has one tier at its max leverage: the table id equalled max leverage for all 292 such markets (2026-10-04).
• "Isolated only" (noCross): a position cannot share margin. "Margin locked" (strictIsolated): in addition, margin cannot be taken out of an open position. The older onlyIsolated flag marks either.
• Growth mode and fee scale are fee settings, shown as set: two documentation pages describe growth mode's effect differently.

The same asset on several venues

Markets on different venues are the same asset when they have the same name and their prices are within 2% of each other.

• The name alone is not enough: main STX was $0.40 and para:STX $848 (2026-10-04). Different names (UBTC on spot, BTC perp) are not matched: no guessing.
• Delisted markets and spot pairs with no trade in 24 hours carry stale prices and take no part.

Totals per exchange

Open interest and volume summed over an exchange's live markets; its share of all builder-exchange open interest.

• Delisted markets take no part. A spot pair counts as live when it traded in the last 24 hours (117 of 330 on 2026-10-04).
• Spot volume sums only pairs quoted in USDC (99.97% of spot volume on 2026-10-04): other pairs are in other tokens, and adding them would mix units.
• Each perp exchange settles in one token, and its prices, volume and open interest are in it. On 2026-10-04 every exchange with live markets settled in USDC; five with none left settle in USDH (flx, vntl, km), USDE (hyna) or USDT0 (cash). Should one of those list markets again, its figures will be in that stablecoin, not USDC.

Changelog

Every change to a definition, dated: a change alters what an embed, a download or a post made before it meant.

• 2026-10-04: Definitions first published here (before the site's public release).