ironflow/hyperliquid-candles-1h

Hyperliquid Hourly Candles

Hourly OHLCV candles for every Hyperliquid perp and spot pair from 25 May 2025 to 31 October 2026: 3,642,453 candles across 799 markets, 466 perps (207 native and 259 HIP-3 builder markets such as xyz:NVDA) and 333 spot pairs.

The busiest hour in the set is 21:00 UTC on 10 October 2025, the October crash: $13.2 billion traded across all markets in one hour, $4.3 billion of it on BTC in 233,527 trades.

Every candle is built from the fills on Ironflow's own Hyperliquid nodes, so delisted markets and HIP-3 markets are in it too. Hyperliquid's own API returns the last 5,000 candles per market and interval; this is the full hourly history in one download.

Get live data

Live Hyperliquid data is on Ironflow. The same candles are on Ironflow's API: GET /v1/candles, any interval from 1 minute to 1 week. No key is needed for the last 24 hours; a free key gives 30 days.

Source and updates

• Source: fills recorded by Ironflow's own Hyperliquid nodes, aggregated per market and UTC hour.
• Checks: compared hour by hour with Hyperliquid's own candleSnapshot (see Notes).
• Updates: refreshed monthly, after each UTC month closes. For anything newer, use the API.
• Coverage: 25 May 2025 to 31 October 2026. Native perps, HIP-3 perps and spot; HIP-4 outcome markets are not included.

Files

One Parquet file per month (UTC): data/YYYY-MM.parquet, one row per market per hour with at least one trade.

Column	Type	Meaning
timestamp	timestamp (UTC)	start of the hour
market	string	Ironflow market symbol, e.g. BTC-PERP, xyz:NVDA-PERP, HYPE/USDC
coin	string	Hyperliquid's name for the market, e.g. BTC, xyz:NVDA, @107
dex	string	HIP-3 builder dex (xyz, para, ...); empty for native perps and spot
market_class	string	perp or spot
open	float	price of the first trade in the hour
high	float	highest trade price
low	float	lowest trade price
close	float	price of the last trade in the hour
volume	float	traded size in the base asset, each trade counted once
trades	int	number of trades

Use

import pandas as pd

c = pd.read_parquet("hf://datasets/ironflow/hyperliquid-candles-1h/data/")
btc = c[c.market == "BTC-PERP"].set_index("timestamp")
print(btc.close.resample("D").last().pct_change().tail())  # daily returns

-- DuckDB: the 10 markets with the most dollar volume in October 2026
SELECT market, sum(volume * close) / 1e9 AS usd_bn, sum(trades) AS trades
FROM 'data/2026-10.parquet'
GROUP BY market
ORDER BY usd_bn DESC
LIMIT 10;

Notes

• An hour with no trades has no row; nothing is forward-filled. Dollar volume is about volume * close.
• Checked against Hyperliquid's candleSnapshot for 20 to 26 October 2026, 168 hours each of BTC, ETH, HYPE, xyz:NVDA and HYPE/USDC: every hour is present; open, close, volume and trade count are identical on the perps. In a few hours the low or high here reaches further (5 of 168 BTC hours, by at most 0.03%), because Hyperliquid's own candles leave some fills out, such as auto-deleveraging during a cascade. HYPE/USDC had 1 to 6 extra trades in 4 of the 168 hours.
• Starts 25 May 2025, the start of Ironflow's checked fill history.
• No duplicate (market, timestamp) rows. Delisted markets are kept up to their last trade.
• Tick-level trades, one row per trade with buyer and seller, are in ironflow/hyperliquid-trades.

Changes

• v1, 1 November 2026. First release.

License and citation

CC BY 4.0. Credit "Ironflow (ironflow.sh)" when you use or share the data.