# Hyperliquid Stock Perps Tracking Error: How Closely Each HIP-3 Market Follows Its Underlying

> We measured 30 days of hourly basis across every live trade.xyz market on Hyperliquid. Here is the tracking error, market by market, with the method and the raw numbers.

*Source: https://hyperliquidguide.com/guides/trading/stock-perps-tracking-error*

A perpetual future named after a stock is not the stock, and the gap between the two has a size you can measure. Funding pulls the perp back toward its reference price over hours rather than instantly, so a crowded market can sit above the reference for a long stretch before the carry drags it home. The oracle feeding the market updates on its own cadence, and the venue where the underlying actually trades shuts overnight and on weekends while the perp carries on. Liquidity does the rest: the same order pushes a thinly traded market much further off its reference than a busy one.

That gap is called basis, and how much it wanders is called tracking error. This page measures both for every live HIP-3 market deployed by [trade.xyz](/guides/trading/hyperliquid-xyz-explained) on Hyperliquid, using 30 days of hourly prints from the exchange's own API.

> **Key takeaway:** Basis tells you where a market is right now relative to its reference. Tracking error tells you how reliably it stays there. A market can carry a small basis today and still be one of the least predictable on the board.

---

## The tracking error of every trade.xyz market

Sorted by 24-hour volume by default. Every column is sortable, and every ticker links to its own market page. Markets listed less than 30 days ago cannot produce a 30-day figure, so they show `n/a` next to the number of days they do cover rather than a partial number dressed up as a full one.

**Trade These Markets With a 4% Fee Discount** — Every market in the table above is live on Hyperliquid right now. The referral discount applies to your first $25M of volume. [Join Hyperliquid](https://app.hyperliquid.xyz/join/Concept211)

---

## How to read basis and tracking error

Take AMD, which was in the top ten by volume when the snapshot was read.

Its mark price was 605.82 and the oracle reference it is anchored to was 605.73. The gap is nine cents on a six-hundred-dollar contract, and the exchange published it as a premium of **+1.54 basis points**. A basis point is 0.01%, so on a $10,000 position that gap is worth $1.54. Positive means the perp was trading above its reference, which is the normal state for a market where more traders want to be long than short.

Now the second number. AMD's **30-day tracking error was 5.1 bps**, and its **mean absolute basis was 3.9 bps**. Read those together and you get the shape of the market rather than a single frame of it: the typical hour over the last month sat about 3.9 bps away from the reference in one direction or the other, and the hour-to-hour scatter around AMD's own average basis had a standard deviation of 5.1 bps. On a $10,000 position that is $5.10 of wobble, one standard deviation.

The funding columns are the third piece. AMD's trailing 24-hour funding annualizes to **+4.7%** and its trailing 7 days to **+7.5%**. Positive means longs were paying shorts, which is what the mechanism does when the perp sits above its reference. Those are annualized rates from a short window, not returns anyone earned; a rate that ran for a day is quoted per year so it can be compared across markets, and it will not stay there.

> **Tip:** Basis is a snapshot and tracking error is a distribution. If you are sizing a single trade, look at the basis. If you are holding for weeks, or running anything automated that assumes the perp and the underlying stay glued together, look at the tracking error.

One thing the number deliberately does not measure: whether the oracle reference is itself correct. Tracking error here is the perp against the price feed its market is anchored to, which is the quantity the exchange's funding mechanism acts on. It is not the perp against the closing print on the listing venue. That distinction matters most for markets whose underlying trades in another time zone, and the [after-hours trading guide](/guides/trading/after-hours-trading-guide) covers what happens to these prices once the cash session shuts.

---

## Which markets diverge most and why

Two cuts of the same data, both computed from the table above.

The left-hand table ranks market types by median tracking error. FX pairs and broad index markets sit at the tight end, single-name equities sit wider, and the range inside the equity group is far larger than the gap between any two groups. The right-hand table splits the same markets into thirds by 24-hour volume. The busiest third tracks more tightly than the quietest third, but the rank correlation printed underneath it is nowhere near -1, which is the honest version of the story: volume explains some of the spread and a long way from all of it.

A few specifics worth pulling out of the rows.

**SP500 was the tightest market on the board at 2.0 bps**, with XYZ100 just behind it. Both are broad index markets with heavy volume, and both stayed within a couple of basis points of their reference for essentially the whole month. If you want a HIP-3 market that behaves like its reference, the index products did that best over this window. The [S&P 500 perpetual guide](/ecosystem/sp500-perpetual-hyperliquid) covers how that market is constructed.

**The two crude oil markets are the clear exception to the volume pattern.** CL and BRENTOIL were among the four highest-volume markets in the study, and they still posted 21.3 and 24.4 bps of tracking error, roughly three times the median and ten times SP500. Heavy trading did not produce a tight basis in these two. We are not going to guess at the reason from a month of hourly prints; what the data says is that oil volume and oil basis behave differently here than equity volume and equity basis do. The [oil futures guide](/guides/trading/trade-oil-futures-on-hyperliquid) and the [commodities guide](/guides/trading/commodities-trading-guide) go into how those contracts are specified.

**The widest markets in the study were the quietest ones.** BIRD at 43.3 bps, GIGADEV at 40.8 and NCLD at 36.7 all sat in the bottom volume bucket. Among names with real turnover, SOFTBANK at 33.9 bps and CXMT at 30.0 bps were the widest, and both reference companies listed outside the United States. Sort the volume column ascending in the table and the pattern shows up immediately: the long tail of the board is where the basis wanders.

None of this makes a wide market a broken one. A perpetual is supposed to drift from its reference and get pulled back; that is the mechanism working, and [funding is the tool that does the pulling](/guides/trading/funding-rates-explained). What the spread means in practice is that the cost of using one of these contracts as a proxy for the underlying is not uniform across the board, and the difference between the tight end and the wide end is more than twenty-fold.

**Check the Live Basis Before You Size a Position** — Every market page on this site carries the live mark, oracle and funding rate, refreshed on load. [Get the 4% Discount](https://app.hyperliquid.xyz/join/Concept211)

---

## What to do with the number

If you are trading a single name for a few hours, tracking error barely registers next to [slippage](/guides/trading/slippage-explained) and the fee. If you are holding for weeks, it compounds with funding into a real cost, and the [90-day funding study](/ecosystem/what-hyperliquid-perps-cost) has the other half of that arithmetic.

What the number is actually good for:

1. **Picking between two similar markets.** If you want index exposure and two products would do, the tighter tracker gives you a price closer to the thing you actually wanted, on both entry and exit.
2. **Sizing.** A market with 30 bps of tracking error can move half a percent against your reference without anything unusual happening. That belongs in the stop, not in the surprise.
3. **Anything automated.** A basis assumption baked into a bot is a number that should come from measurement. The [API guide](/guides/trading/hyperliquid-api-guide) covers reading `metaAndAssetCtxs` and `fundingHistory` yourself if you would rather run your own window.

For the broader picture of what is listed and how the contracts work, the [equity perps guide](/guides/trading/equity-perps-guide) and [non-crypto perps reference](/guides/trading/non-crypto-perps-on-hyperliquid) cover the full roster. To start trading any of them, open [app.hyperliquid.xyz](https://app.hyperliquid.xyz/join/Concept211) and switch to the trade.xyz markets from the asset selector. The [live HIP-3 liquidity screener](/tools/hip-3-liquidity) shows depth and spread on the same markets in real time.

---

## Methodology

**Endpoints.** Everything on this page is read from the public Hyperliquid API at `api.hyperliquid.xyz/info`. The snapshot row for each market, mark price, oracle price, current premium, hourly funding, open interest and 24-hour volume, comes from `{"type":"metaAndAssetCtxs","dex":"xyz"}`. The hourly history comes from `{"type":"fundingHistory","coin":"xyz:","startTime":}`.

**Sampling time.** The snapshot was read at 11:55:32 UTC on 22 September 2026. The history window is the 30 days from 23 August 2026 to 22 September 2026, in UTC.

**What `premium` is.** Each `fundingHistory` row carries a `premium` field, which is Hyperliquid's own published basis of the mark price to the oracle price for that hour, as a decimal. Multiplying it by 10,000 gives basis points. We use that field rather than recomputing the gap from the two prices, because it is the quantity the exchange's own funding calculation acts on. The data file stores both: `basisBps` from `premium`, and `basisFromPricesBps` recomputed as `(mark - oracle) / oracle`. The two agree closely on most markets and diverge a little on some, because the recomputed version uses the mark and oracle from a single instant while `premium` is the exchange's figure for the hour. Where they disagree, `premium` is the one shown in the table.

**Tracking error.** For each market, the sample standard deviation of the hourly `premium` series across the window, using the n minus 1 denominator, multiplied by 10,000 to give basis points. Mean absolute basis is the mean of the absolute value of the same series, in the same units. Both describe the market's behavior over the window and neither predicts the next one.

**Funding.** The 24-hour and 7-day funding columns sum the hourly funding rates over the trailing window and annualize by multiplying by 8760 divided by the number of hours in that window, then by 100. They are annualized rates, not realized returns, and a short window annualized produces a large-looking number by construction.

**Coverage and the `n/a` rule.** `fundingHistory` returns at most 500 rows per call, roughly 20.8 days of hourly prints, so the collection script pages through each market until the window is covered. A market qualifies for a 30-day figure only if its series carries at least 718 of the 720 possible hourly prints. Any market short of that shows `n/a` in the tracking error and mean absolute basis columns, keeps its real day count in the last column, and is excluded from every median and range on this page. On this run, 108 of the 123 markets in the trade.xyz universe were live, and 101 of those 108 had a full series. The seven that did not were listed inside the window.

**Delisted markets.** Any market flagged `isDelisted` in the API response is dropped before anything is computed.

**Update cadence.** The page is refreshed by re-running `scripts/collect-tracking-study.js` against a rolling 30-day window. Every figure carries the window it was measured over, so a quoted number is a point-in-time measurement rather than a standing claim. The collection timestamp printed above the table is the authoritative one.

**Reuse.** You may republish these figures with attribution and a link to https://hyperliquidguide.com/guides/trading/stock-perps-tracking-error.

**Start Trading Stock Perps on Hyperliquid** — Cash-settled, USDC margin, no expiry, and 24/7 on every market in the table above. [Join Hyperliquid](https://app.hyperliquid.xyz/join/Concept211)

30-day tracking error and basis for selected Hyperliquid HIP-3 markets, window 2026-08-23 to 2026-09-22. Tracking error is the sample standard deviation of the hourly mark-versus-oracle premium, in basis points. Source: Hyperliquid API, via Hyperliquid Guide (https://hyperliquidguide.com/guides/trading/stock-perps-tracking-error)MarketTypeBasis (bps)30d tracking error (bps)Mean abs. basis (bps)SP500Index-1.872.02.0XYZ100Index+2.462.21.6MSFTEquity-2.642.72.3SILVERCommodity+6.223.12.7AAPLEquity+3.323.23.2NVDAEquity+3.133.72.8AMDEquity+1.545.13.9CLCommodity-3.6721.312.9BRENTOILCommodity+3.2324.414.0CXMTEquity-29.4830.019.2SOFTBANKEquity-31.2533.925.1BIRDEquity+77.9243.333.4`}
/>