KerbTape

Methodology

Where the numbers on the board come from, what they mean, and what they can’t tell you.

Data source

Every price on this site comes from Hyperliquid’s public API (api.hyperliquid.xyz). The markets are HIP-3 perpetual futures listed by the xyz deployer, which track stocks, indices and commodities and keep trading 24/7, including when their home exchange is shut.

We don’t use or show data from NYSE, Nasdaq, CME, the Korea Exchange or the Tokyo Stock Exchange. Exchange calendars (holidays, early closes, daylight saving) decide when each market counts as closed. Markets use Hyperliquid’s own names and tickers (for example S&P 500, ticker S&P500; Japan 225, ticker JP225).

Definitions

Last close
The market’s price on Hyperliquid at the minute its home exchange closed (for example 16:00 ET on Friday for US stocks, 15:30 KST for Korea, 15:30 JST for Japan, 17:00 ET for CME energy and metals). How we get it is explained below; it is a proxy for the official close, not the exchange’s own closing price.
Now
The mid price on Hyperliquid when the order book has both sides, otherwise the mark price.
Gap
gap = (now ÷ last close − 1), shown as a percentage with its sign. A gap is only shown while the home exchange is closed; open markets show their 24-hour change instead. The last close is the home exchange’s regular-session (cash) close, so the gap is measured from the cash close, not from the last real price.
Gap vs the last real price
For several markets real prices keep coming after the cash close: US stocks trade after hours and overnight until 20:00 ET on Friday, Korean stocks have an after-hours session until 20:00 KST, and Korea 200 and Japan 225 follow index futures that trade into the night (Korea 200 until 05:50 KST on Saturday). Until then Hyperliquid’s oracle follows those real prices. So part of a gap can be a move that already traded on a real market on Friday evening; only what happens after real pricing stops is weekend price discovery. Where we captured the price at that moment, the board shows now ÷ last real price − 1 as “since real pricing stopped” next to the gap. CME energy and metals have no such evening session, so for them the two are the same.
Countdowns: when real pricing resumes
The countdowns show when Hyperliquid’s oracle goes back to following real prices, which is when the weekend price meets a real market again. For most markets that is the home exchange’s open. US stocks come back earlier, in the overnight session (Sunday 20:05 ET), and US indices, Korea 200 and Japan 225 come back with their index futures on Sunday evening (US time). Those countdowns say “real pricing resumes” and name the later cash open next to it, for example “real pricing resumes Sun 20:05 ET (overnight session) · cash open Mon 09:30 ET”. The terminal uses the same times.
Liquidity
Graded from Hyperliquid’s 24-hour volume and open interest for the perpetual, with the same thresholds and words as the terminal and the API. A thin book has under $250k of 24-hour volume and under $5M of open interest: a few small trades can move its price and funding, so its gap says less. Moderate liquidity means under $1M of volume and under $25M of open interest. Anything above that is liquid and isn’t marked. Rows show a small “thin” or “moderate” tag.
Gap bar
The vertical line in every row is the row’s reference: the last close, or the last real price where the row leads with the move since real pricing stopped. The bar shows the big figure: it grows right when it is positive and left when it is negative. The scale is compressed (logarithmic above 0.25%) so small gaps stay visible, and the bar fills completely at ±5%.
Reopen reference
The median oracle price from 5 to 10 minutes after the home exchange reopens. Using a few minutes of median rather than the first print avoids the opening auction’s noise. If we didn’t record the oracle then, we use Hyperliquid trade prices from just after the reopen instead.
24h volume, open interest, funding
As reported by Hyperliquid for the perpetual. Funding is the hourly rate; the yearly figure is that rate × 24 × 365 and only a rough guide.

How last closes are derived

Each last close comes from Hyperliquid’s own prices, by the first rule below that applies. The board says when a close is estimated.

  1. Oracle at the close. When our collector was running at the close, we take Hyperliquid’s oracle price for the market from the snapshot nearest the closing minute (up to 2 minutes before or 90 seconds after). While the home exchange is open the oracle tracks it, so this is the closest thing to the official close that Hyperliquid publishes. This is the only method that isn’t marked as an estimate.
  2. Estimated from trades. When we have no oracle snapshot at the close (for example, the collector started after it), we estimate it from Hyperliquid trades in the minutes before the close:
    • Liquid markets (trades in at least 8 of the last 10 one-minute periods before the close): the last trade at the close. On busy markets that is the trade nearest the closing auction, and in our checks against official closes it was closer than an average.
    • Thin markets: a volume-weighted average of trade prices over the last 10 minutes before the close, skipping minutes with no trades. If fewer than 3 of those minutes had trades, the window widens to 30 minutes.
    • If there were no trades at all in that window, we fall back to the last trade before the close.
    These closes are marked est. on the board and say “Last close estimated from Hyperliquid trades” on the market page.

Limits of the estimate. On busy markets (S&P 500, the big US stocks, gold, oil) a trade-based close usually lands within a few hundredths of a percent of the official close. On thin markets, such as Japan 225 and Korea 200, several minutes can pass without a trade and the few trades that happen may be small and away from the exchange’s closing auction. There an estimated close can differ from the official exchange close by tens of basis points (a few tenths of a percent), and the gap inherits that error. Treat small gaps on estimated, thin markets with caution.

Delays and missing prices

Scorecard method

For each market and each closure we record the last close, the weekend price (5 minutes before the reopen) and the reopen reference. We then compare two errors: weekend price vs reopen, and last close vs reopen. The weekend price “wins” when its error is smaller.

Each week we publish the win rate, the median absolute error of both and how much the weekend price reduced it, and whether the direction was right on moves of at least 0.25% and 1%, split by asset class. A running 12-week table sits underneath.

Limitations

Privacy and referrals

We count page views and Trade clicks as daily totals by referral code and market, with no IP addresses, user agents or cookies. When a trader who arrived through a partner’s referral link approves the builder fee, we record their wallet address against that code, so the partner can be credited with the volume it brings (read from Hyperliquid’s public fills). We publish totals per code only, never addresses. More on the About page.

Not investment advice

KerbTape describes where prices are. It doesn’t tell you what to buy or sell, and nothing here is a forecast. Perpetual futures use leverage and you can lose all of your margin.