Kumo Research · Forex
Forex on-chain: do Hyperliquid's EUR/USD, GBP/USD and USD/JPY perps track the real currency market?
Currency trading has always been the domain of banks and FX brokers. Since 2026, perpetual futures on the euro, the pound and the yen have also traded on Hyperliquid through trade.xyz, settled in USDC and open around the clock. We checked how closely they follow the real exchange rates and how they behave when the FX market is closed.
By Alex Wu · Published · 6 min read
Key findings
- The euro perp tracked EUR/USD with a median gap of just 1.75 basis points (0.0175%), within 6 bp in 95% of hours, and its hourly moves had a 0.89 correlation with the real rate.
- The yen perp tracked USD/JPY within a median 3.6 bp and moved almost exactly with it (correlation 0.95), but traded on average 2.8 bp below the market rate.
- The pound perp, launched in May 2026, was the loosest: a median gap of 3.9 bp, 14 bp at the 95th percentile and a much weaker hourly correlation (0.46).
- On weekends, when the real FX market is closed, the perps barely moved: their average 48-hour move from Friday to Sunday evening was a quarter to a third of a normal weekday 48 hours.
- Funding pointed the same way as traditional FX carry: longs paid to hold the euro perp (+2.9% a year) and were paid to hold USD/JPY (−4.7% a year, shorts paying).
A new way to trade currencies
Most retail FX trading happens through brokers who quote prices from bank liquidity. On-chain perpetual futures offer a different route: anyone with USDC can trade a perp that is meant to follow EUR/USD, GBP/USD or USD/JPY, with no broker in between and no market closure at the weekend. Hyperliquid lists these markets through trade.xyz as xyz:EUR, xyz:GBP and xyz:JPY (the last one quoted as yen per dollar, like USD/JPY).
The obvious question is whether they actually track the real exchange rate. In our stock-perps study we found that Hyperliquid's stock perps stayed within a few hundredths of a percent of the real share price. Currencies are a harder test: they move in much smaller steps, so even a small tracking gap matters.
How we measured it
We compared each perp's hourly closing price with the market exchange rate at the same moment (from Yahoo Finance's hourly FX data) on every weekday hour from April 2026 – May for the pound – to 2 November 2026: about 9,700 comparisons in total. The gap is shown in basis points; for EUR/USD, 1 bp is about 0.0001 – one "pip".
How close is close?
| Pair (Hyperliquid market) | Hourly comparisons | Average gap | Median absolute gap | 95th pct absolute gap | Hourly return correlation | Funding, annualised (longs pay +) |
|---|---|---|---|---|---|---|
| EUR/USD (xyz:EUR) | 3456 | +0.5 bp | 1.7 bp | 6.0 bp | 0.89 | +2.9% |
| GBP/USD (xyz:GBP) | 2795 | -0.6 bp | 3.9 bp | 14.3 bp | 0.46 | -3.2% |
| USD/JPY (xyz:JPY) | 3437 | -2.8 bp | 3.5 bp | 8.0 bp | 0.95 | -4.7% |
Weekday hours, 8 April (GBP: 18 May) – 2 November 2026. Gap = perp ÷ market rate − 1 at the same hour. Funding averaged from April 2026 (GBP from May).
- EUR/USD was the best: a median gap of 1.75 bp – under two pips – and within 6 bp in 95% of hours. The perp's hourly moves had a 0.89 correlation with the real rate.
- USD/JPY tracked within a median 3.6 bp and moved almost in lockstep with the market (correlation 0.95). On average it sat 2.8 bp below the market rate – a small but consistent discount.
- GBP/USD, the newest market, was the loosest: a median gap of 3.9 bp, 14 bp at the 95th percentile, and a much weaker hourly correlation of 0.46. Its hourly price was noisier around the real rate – typical for a younger, less liquid market.
For comparison, the typical spread on EUR/USD at a retail FX broker is often around one pip or less in liquid hours. The euro perp's median tracking gap is of a similar size.
By trading session
How closely on-chain FX tracked the real exchange rate
Median gap between the Hyperliquid perp and the market rate, by session, basis points
- EUR/USD
- GBP/USD
- USD/JPY
Source: Hyperliquid trade.xyz hourly closes vs Yahoo Finance hourly FX rates, April – November 2026 · Chart: Kumo Research, kumocast.io
Tracking was steady across the day. EUR/USD stayed within about 1.6–1.9 bp in every main session and widened only slightly in the quiet 5–7 pm New York window (2.4 bp), when real FX liquidity is thinnest. USD/JPY was equally steady in the Asian, London and New York sessions. The pound perp was looser in every session, not just one, which points to the market's own depth rather than a timing problem.
Weekends: open, but quiet
The real FX market closes from about 5 pm New York time on Friday to 5 pm on Sunday. The perps keep trading. We measured how much each perp moved over that 48-hour weekend window and compared it with an ordinary 48 hours during the week.
Weekends on-chain are quiet
Average 48-hour move of the perp: weekend (Fri 5 pm – Sun 5 pm NY) vs weekdays, basis points
- Weekend
- Weekday
Source: Hyperliquid trade.xyz hourly closes, April – October 2026 · Chart: Kumo Research, kumocast.io
On average the weekend move was 12 bp for the euro perp against 36 bp over a weekday 48 hours, 9 bp against 46 bp for the yen, and 16 against 41 bp for the pound. With no bank market to price against and little news, the perps mostly drifted sideways until the real market reopened on Sunday evening. This matches what we saw with stock perps: weekend prices carry little information until the traditional markets reopen.
Does funding mirror FX carry?
In traditional currency markets, holding a currency with a higher interest rate than the one you borrow earns "carry", and the reverse costs money. Perps have no interest rates, but they have funding payments between longs and shorts – and those can reflect the same pressure.
Since April 2026, longs on the euro perp paid an average of +2.9% a year in funding, with payments going the other way in only 3% of hours. On the yen market (USD/JPY), funding averaged −4.7% a year: shorts paid longs about half the time, and the payments they made were larger. That is the same direction as the classic carry trade of recent years, in which being long the US dollar against the euro or the yen earned interest. The pound market averaged −3.2% a year, with longs paid.
We cannot prove that carry is the cause – funding also reflects how many traders want to be long or short – but the direction is consistent with how these currency pairs are traditionally financed.
What this means
- Hyperliquid's euro and yen perps have tracked the real exchange rates very closely – within a couple of pips most of the time.
- The pound perp, launched later, has been noisier around the real rate.
- On weekends the perps trade but barely move; the real price discovery still happens when the FX market is open.
- Funding costs on these markets have pointed in the same direction as traditional FX carry, which matters for anyone holding a position for weeks.
For live EUR/USD, GBP/USD and USD/JPY levels and volatility ranges, see our EUR/USD, GBP/USD and USD/JPY pages, and our study on when the forex market moves.
Method, data and limits
Data: hourly closes of Hyperliquid trade.xyz perps xyz:EUR, xyz:GBP and xyz:JPY (from 8 April 2026; GBP from 18 May 2026) and the opening rate of Yahoo Finance's hourly FX bars at the same timestamp, to 2 November 2026, weekdays only. Hyperliquid hourly funding history for the same markets.
Measures: gap = perp ÷ market rate − 1 in basis points; hourly return correlation between the perp and the market rate; weekend move = |log change| from Friday 5 pm to Sunday 5 pm New York time, compared with 48-hour windows starting at 5 pm Monday to Wednesday. Funding annualised as mean hourly rate × 8,760.
Limits: Yahoo FX rates are indicative mid-quotes, not executable prices, which can add noise to the gaps (especially for GBP); seven months of data; one venue. Not investment advice.
Download the data (CSV, CC BY 4.0)
How to cite
Kumo Research (2026). "Forex on-chain: do Hyperliquid's EUR/USD, GBP/USD and USD/JPY perps track the real currency market?". Kumocast. https://www.kumocast.io/research/forex-onchain-perps-vs-real-fx
This study describes past market data. It is not investment advice and past behaviour does not guarantee future results.