kumocast

Kumo Research · Stock perps

Do on-chain stock perps track the real stock? Hyperliquid and OKX vs Nasdaq prices, hour by hour

Perpetual futures on US stocks now trade around the clock on crypto venues such as Hyperliquid (through trade.xyz) and OKX. They have no expiry and no direct link to the stock exchange – the price is held in line only by traders and funding payments. We checked how closely they actually follow the real share price.

By · Published · 6 min read

Key findings

  1. During regular US trading hours, Hyperliquid stock perps stayed within a median of 3.1 basis points (0.031%) of the real stock price, and within 8.9 bp in 95% of hourly checks.
  2. Tracking was almost as tight in pre-market and after-hours trading (median 3.4 and 3.3 bp), even though those sessions are much thinner on the stock exchange.
  3. OKX's USDT-margined stock swaps showed a steady offset of about +11 bp in every session – a near-constant gap consistent with pricing in USDT rather than dollars – but moved with the stock just as closely around that offset.
  4. Big stocks tracked best: Nvidia and Microsoft had median gaps of 2.3–2.5 bp; Coinbase, the most volatile stock in the group, had the widest at 5.1 bp.
  5. The perps trade all week, but the action follows Wall Street. Regular hours are 21% of the clock yet carry 63% of the perps' price movement and 55% of their dollar volume; weekends are 34% of the clock but only 6% of the movement.

What are stock perps?

A perpetual future ("perp") is a contract that tracks the price of an asset but never expires. Perps on Bitcoin and Ether have been the most traded crypto products for years. Since 2025, crypto venues have also listed perps on US stocks, indices, gold and currencies. On Hyperliquid they trade through trade.xyz's markets (quoted in USDC); OKX lists USDT-margined stock swaps.

There is no share behind a stock perp. Its price stays near the real stock because of two mechanisms: arbitrage traders who buy the cheaper one and sell the more expensive one, and a funding rate that charges whichever side is pushing the perp away from its reference price. Whether those mechanisms actually work is an empirical question. If the perp drifts, people trading it are not getting the exposure they think they are.

How we measured it

We took the hourly closing price of eight stock perps – Tesla, Nvidia, Apple, Microsoft, Amazon, Meta, Alphabet and Coinbase – and compared each one with the real stock price at exactly the same minute, using five-minute prices from US exchanges including pre-market and after-hours trading. The sample covers 10 August to 2 November 2026: about 7,400 matched hourly comparisons per venue on weekdays.

The gap is shown in basis points (bp). One basis point is 0.01%, so a 3 bp gap on a $300 stock is 9 cents.

Regular hours: within a few cents

How closely Hyperliquid stock perps tracked the real stock

Median gap between perp price and NYSE/Nasdaq price during regular hours, basis points (1 bp = 0.01%)

  • NVDA2.3 bp
  • MSFT2.5 bp
  • TSLA2.9 bp
  • AMZN2.9 bp
  • AAPL3.0 bp
  • META3.3 bp
  • GOOGL3.8 bp
  • COIN5.1 bp

Source: Hyperliquid trade.xyz hourly closes vs Yahoo Finance 5-minute prices, 10 Aug – 2 Nov 2026 · Chart: Kumo Research, kumocast.io

During regular trading hours (9:30 to 16:00 New York time), Hyperliquid stock perps tracked the real stock with a median gap of 3.1 bp. In 95% of hourly checks the gap was below 8.9 bp, and it went above 20 bp (0.2%) less than 1% of the time.

The tightest tracking was in the largest and most liquid names: Nvidia (2.3 bp) and Microsoft (2.5 bp). Coinbase, which is itself a volatile, crypto-linked stock, had the widest median gap at 5.1 bp – still about five cents on a $100 share.

Hyperliquid perps also showed a small upward bias: on average they traded 1.6 bp above the stock during regular hours. That is consistent with more traders wanting to be long than short, and it is the reason funding usually flows from longs to shorts (we look at funding costs in a separate study).

Pre-market and after-hours

Perp price vs the real stock price, by venue and session (8 stocks)
VenueSessionHourly comparisonsAverage gapMedian absolute gap95th pct absolute gap
Hyperliquid (trade.xyz, USDC)After-hours (16:00–20:00 ET)1697+2.0 bp3.3 bp10.6 bp
Hyperliquid (trade.xyz, USDC)Pre-market (4:00–9:30 ET)2339+1.8 bp3.4 bp10.4 bp
Hyperliquid (trade.xyz, USDC)Regular (9:30–16:00 ET)3360+1.6 bp3.1 bp8.9 bp
OKX (USDT-margined)After-hours (16:00–20:00 ET)1697+11.0 bp10.0 bp23.6 bp
OKX (USDT-margined)Pre-market (4:00–9:30 ET)2339+11.9 bp11.6 bp22.6 bp
OKX (USDT-margined)Regular (9:30–16:00 ET)3360+11.1 bp10.7 bp21.0 bp

Gap = perp price ÷ stock price − 1 at the same minute, on weekdays, 10 August – 2 November 2026. Stocks: TSLA, NVDA, AAPL, MSFT, AMZN, META, GOOGL, COIN.

The extended sessions are where many people expected perps to drift, because the stock itself trades thinly. In practice tracking was nearly as good: a median gap of 3.4 bp in pre-market (4:00–9:30 ET) and 3.3 bp after hours (16:00–20:00 ET), with 95% of checks within about 10.5 bp.

The few large gaps we found all came in after-hours trading, when the stock exchange price is based on very few trades. The largest was 5.8% for Alphabet at 17:00 ET on 18 August 2026; three others exceeded 0.5%. In those moments it is not obvious which price is "wrong" – the perp, with more continuous trading, may simply have been ahead of a thin after-hours print.

OKX: same tracking, different starting point

OKX's stock swaps showed a different pattern. Their gap to the stock was about +11 bp in every session – pre-market, regular and after-hours alike – with a median absolute gap of 10–12 bp. The stability of that offset suggests a systematic reason rather than poor tracking: OKX contracts are priced in USDT, and a USDT price slightly below one dollar, together with OKX's funding settings, would produce a steady small premium. Around that offset, the OKX contracts moved with the stock as closely as Hyperliquid's.

For a trader, the practical point is that OKX prices for these contracts should not be compared one-for-one with the stock quote: the offset is small, but it is persistent.

When do stock perps actually trade?

Stock perps are open 24 hours a day, seven days a week. That raises the question of how much real trading happens while Wall Street is closed. We split the clock into four parts and measured, for each, its share of hours, its share of total price movement (variance of hourly returns) and its share of dollar volume, from mid-April to November 2026.

Open all week, busy only when Wall Street is

Share of clock time, price movement and dollar volume of 8 stock perps, by US market session, %

0%20%40%60%Regular hoursPre-market & after-hoursOvernight (8 pm – 4 am ET)Weekend
  • Share of hours
  • Share of price movement (variance)
  • Share of dollar volume

Source: Hyperliquid trade.xyz hourly data, 15 Apr – 2 Nov 2026 · Chart: Kumo Research, kumocast.io

  • Regular US hours are only 21% of the week but carry 63% of all price movement and 55% of the dollar volume.
  • Pre-market and after-hours are 27% of the week, with about a quarter of movement and volume. This is where earnings reactions happen.
  • Overnight (8 pm – 4 am New York time on weekdays) is 19% of the clock but only 6% of movement and 11% of volume.
  • Weekends are a third of all hours, yet only 6% of price movement and 8.5% of volume.

The perps are mostly used as a way to trade US stocks in US hours from anywhere, with crypto collateral – and only secondarily as a way to trade when the stock market is closed. What happens to weekend prices, and whether they predict Monday's opening, is the subject of our next stock-perps study.

What this means

  1. Hyperliquid's stock perps have tracked the real share price very closely: about 3 bp in the median and under 9–11 bp in 95% of hours, across all US sessions.
  2. OKX's USDT-margined stock swaps carry a steady offset of about 11 bp but follow the stock just as closely around it.
  3. Most activity in stock perps still happens while the US market is open; weekends and overnight hours are quiet.

For live levels and volatility ranges of these markets, see our stock pages, such as Tesla, Nvidia and Coinbase.

Method, data and limits

Data: hourly closes of Hyperliquid trade.xyz perps (xyz:TSLA, NVDA, AAPL, MSFT, AMZN, META, GOOGL, COIN) and OKX USDT swaps (e.g. TSLA-USDT-SWAP); five-minute prices from Yahoo Finance including pre- and post-market trading. Comparison window 10 August – 2 November 2026 (weekdays); session shares use Hyperliquid hourly data from 15 April 2026.

Matching: each perp hourly close (at hh:00 UTC) is compared with the stock's five-minute bar closing at the same minute. Gap = perp ÷ stock − 1, in basis points. Sessions are defined in New York time.

Session shares: variance share = sum of squared hourly log returns in a session ÷ total; dollar volume = base volume × close.

Limits: extended-hours stock prices can be thin; Yahoo data can contain occasional bad prints; perps and stock quotes are not executable at the same size. Not investment advice.

How to cite

Kumo Research (2026). "Do on-chain stock perps track the real stock? Hyperliquid and OKX vs Nasdaq prices, hour by hour". Kumocast. https://www.kumocast.io/research/stock-perps-vs-real-stock-price

This study describes past market data. It is not investment advice and past behaviour does not guarantee future results.

Live data for these markets

More from Kumo Research

Alex Wu, Independent market analyst

About the researcher

Alex Wu

Crypto, DeFi, perps and FX · driven by data

Alex designed the Kumocast model behind these signals, sets its rules, reviews its accuracy record and updates the methodology whenever the model changes. Signals are refreshed every 30 minutes with live market data.