# How we rate multi-asset perpetual venues

Cross-Asset Desk Research; Updated 24 September 2026

One scoring sheet serves the whole domain. Version 1.1 of it was cut on
23 September 2026 and fixed before the sheet below was recomputed, so the weights
answer the question this desk asks rather than the table they go on to produce; what moved in this
version, and why, is set out under [what changed at version 1.1](#what-changed-at-version-1-1).
Three decisions sit behind any row you read here, and they are
made in three different places: whether a venue belongs on a page at all is a question of the page's
subject — a gold market, a share listing, a venue somebody is walking away from; what the venue scores is
a question of the weights below; and [what puts one venue above another](#what-puts-one-venue-above-another)
is a question of the column that page turns on. The scale stops at 100.

## What is being weighed

| Criterion | Weight | What the criterion measures | How it becomes a score | Marks it actually spanned |
|---|---|---|---|---|
| What one account reaches | 32 | The reach of one account — perpetual contracts as the venue's own documentation counts them, and how many asset classes outside crypto those contracts cover. | Two halves, and not equal ones. Six points sit on the size of the list, paid along a curve whose ceiling arrives at 500 contracts: a hundred leaves the mark at four and a half, twenty at under three, so no crypto board however long takes this criterion on its own. The other four come flat — a point for each class beyond crypto, metals, energy, currencies, equities, indices, pre-listing lines — and the fourth one ends the count. Ten needs both halves. | 7.8 to 10.0 across the six venues on the register, and the same 7.8 to 10.0 wherever a page carries five of them |
| Cost of getting in and out | 28 | What it costs to be in and out again before the market has moved at all — the entry-tier taker rate on a BTC perpetual, charged once on the opening fill and once on the closing one, read in basis points. | On the ten-thousand-dollar round trip this desk prices, a basis point is a dollar. The first two cost nothing in points — a round trip at two or under keeps all ten — and every dollar after that takes off half a point, so twelve basis points scores five and at twenty-two the column is down to zero. | 3.0 to 10.0 across the six venues on the register, the widest span of the five criteria |
| Custody of the margin | 18 | What can be done with the margin while a position is open, and by whom — answered out of what the venue has written down and nothing else, from settlement notes to withdrawal terms to audit reports. | every trade lands on a chain, in a rollup's batch or under a proof of validity: 4; the chain is given a balance at intervals and never the trades that moved it: 2; the terms give the trader a withdrawal the operator cannot stand in front of: 3; someone outside the venue has audited the code that holds the margin: 2; the trader arrives with a wallet and leaves no account behind: 1 | 5.0 to 10.0 across the six venues on the register, and unchanged on the pages that carry five |
| How the venue has held up | 12 | Read off the record rather than off a venue's account of itself — the audits it can show, the bounty it keeps open, and whether the past 24 months hold a loss of customer money. | an audit by an outside firm is two points, and the sheet stops counting at the second: 4; a standing bounty for whoever finds the next flaw: 3; two years of trading with none of the customers' money lost: 3 | 4.0 to 10.0 across the six venues on the register, on marks that fall into three groups rather than a spread |
| The book behind the quote | 10 | What is actually standing behind a printed price — resting bids and offers on the BTC perpetual, counted together where they sit inside a tenth of a percent of mid, and filed as the median of the desk's polls. | Count the zeros. Ten thousand dollars resting behind the quote is the starting mark and earns nothing, as does anything under it; each zero added to that figure is worth two more points, so a book ten times the size marks two, a hundred times four, a thousand times six. Ten is where the scale stops paying. | 5.2 to 6.6 across the six venues on the register, the narrowest span of the five criteria |


The final column belongs beside the weight because the two only mean something together. A criterion
can lift one venue over another only as far as their marks are apart, so a large share of the hundred
laid over marks that bunch does less than a small share laid over marks that scatter. Across the six
venues on the register, cost holds twenty-eight points and spans 7.0 marks, the widest of the five;
the book holds ten and spans 1.4, the narrowest; coverage, the heaviest criterion at thirty-two,
spans 2.2. The cause of that last figure sits in the coverage rule and in the register together: four
of coverage's ten points go to asset classes beyond crypto and are capped at four, and no venue is on
the register with fewer than four, so every venue on the sheet takes that half of the criterion in
full. What is left to separate them is the raw market count, which on all six is mostly crypto. The
criterion this domain leads with therefore separates its own table by less than cost, custody or
standing do, and only the book separates it by less still. That is stated rather than corrected. The
weights did move once, at version 1.1, and the reason is set out below and is not this one: a weight
moved to widen a span, or to lift a name up the table, stops being a rule and becomes a preference.

Every criterion is marked out of ten on its own evidence, and the weighted result is then stretched onto
a scale ending at 100. Coverage is the one that needs unpacking: it adds a curve over
the number of perpetual markets, which reaches its six-point ceiling once a venue lists 500 of them, to
one point for each asset class the venue runs beyond crypto, up to four. A venue with thirty markets
across four asset classes and a venue with five hundred crypto pairs can therefore finish level on
coverage and mean completely different things by it, which is why the asset-class count is printed in
its own column rather than buried in the total.

## What changed at version 1.1

The first version of this sheet ran reach 30, in and out 25, margin 20, book 15, standing 10. This one
runs reach 32, in and out 28, margin 18, standing 12, book 10. Sixty of the hundred now sit on the two
questions a reader arrives with — how much of a portfolio one account can actually hold, and what it
costs to move in and out of it. Those are the two this desk was opened to answer; in the first version
they held fifty-five of the hundred between them, which understated how much of the answer they are.

The book gave up the five points. It is read on the BTC perpetual and nowhere else, since BTC is the
only contract all six venues run, and a thick BTC book says nothing about the gold, crude or equity
book quoted beside it — which is the book a reader of these pages is about to trade. A criterion that
answers a question the site is not asking should not outweigh one that answers the question it is.
Standing passed it for the opposite reason: an audit, an open bounty and a clean two years apply to
the whole balance rather than to one contract on it. Margin gave up the two points that move sits on,
and keeps the larger share of the pair, because what a venue may do with the collateral today still
outranks what it has come through.

What did not happen: the weights were not tried against the table first. They were cut from the list
above, fixed, and the sheet recomputed once. The sequence came out as it had before — Lighter,
Extended, Aster, Hyperliquid, EVEDEX, Aevo — with the totals a point or two apart from version 1.0,
and had it come out differently it would have been published differently.

## Why the hours sit outside the formula

Because only three venues state them for everything they list. EVEDEX says all its markets trade 24/7;
Ostium and gTrade document the session each of their non-crypto classes follows. Two more — Lighter
and Extended — state hours for part of a listing and not for the venue, and the three remaining names
on the register say nothing at all. Scoring statements of different scopes against each other, and
marking silence as though it were a stated policy, would both manufacture a fact.

So the hours are printed in the tables rather than folded into a score, quoted from each venue's own
pages, and one page — [where gold trades](/trade-gold-with-crypto) — is put in order by them, because
that is the question the page exists to answer.

## The sheet, venue by venue

The sheet stands as of 18 September 2026 and is recomputed from the columns beside it by the arithmetic
set out above. The five criterion columns are named by the short form the tables carry and run in the
order of the weights — reach, in and out, margin, standing, book — with the full wording of each in
the first table on this page. Round trip doubles the entry-tier taker rate; depth is the middle
reading of the BTC book polls, counting orders that rested inside 10 basis points of the mid.

| Venue | Score | Reach | In and out | Margin | Standing | Book | Round trip | Depth | Asset classes |
|---|---|---|---|---|---|---|---|---|---|
| Lighter | 90 | 9.2 | 10.0 | 10.0 | 7.0 | 6.4 | 0.0 bps | $16.5 million | crypto, stocks, FX, commodities, indices, pre-IPO, bonds |
| Extended | 85 | 9.6 | 8.5 | 7.0 | 10.0 | 5.7 | 5.0 bps | $7.4 million | crypto, stocks, FX, commodities, indices, pre-IPO |
| Aster | 82 | 10.0 | 7.0 | 7.0 | 10.0 | 6.3 | 8.0 bps | $13.8 million | crypto, stocks, FX, commodities, indices, pre-IPO |
| Hyperliquid | 76 | 9.6 | 6.5 | 7.0 | 7.0 | 6.0 | 9.0 bps | $10.1 million | crypto, stocks, FX, commodities, indices |
| EVEDEX | 67 | 7.8 | 6.5 | 5.0 | 7.0 | 6.6 | 9.0 bps | $19.1 million | crypto, commodities, FX, stocks, indices, pre-IPO |
| Aevo | 58 | 8.5 | 3.0 | 7.0 | 4.0 | 5.2 | 16.0 bps | $3.8 million | crypto, stocks, commodities, FX, pre-IPO |

## What puts one venue above another

The page's subject decides who is on it. What decides the sequence is whatever the reader came to find
out: a page built on one documented fact — the hours a market keeps, the rate it charges — is sorted by
that fact, and every other page is sorted by the total.

| Page | Ordered by |
|---|---|
| Home, Oil, Currencies, One account, Ostium alternatives, Stocks | The five criteria weighted and added, largest total at the top |
| Fees | What a $10,000 entry and exit takes in taker fees on day one, cheapest first |
| Gold | How far a venue's statement on hours reaches, the ones that set them venue-wide first |


The total stays in the table either way, where it settles two venues that land on the same figure.
Nothing about the criteria, the weights or the arithmetic shifts between pages, so a venue's number is
the same wherever you meet it here.

## Who is on the sheet, and who is not

Venues listing perpetual contracts on assets beyond crypto — metals, energy, currencies, equities or indices — against stablecoin collateral. One line draws the register, and it is read off each venue's own asset-class list rather than off anything this desk measures: a venue is scored here only when its documentation lists at least four asset classes that are not crypto. Below four, one margin balance is a crypto book with a sideline attached, which is a different instrument from the one these pages are about. Nine of the sixteen venues followed here clear that line, and six of those run the resting order book the liquidity criterion has to read: EVEDEX, Extended, Aevo, Hyperliquid, Lighter and Aster. Those six are the register. Ostium and gTrade clear the line as well and are listed with their facts and no score, because they price against an oracle and fill from a pool rather than from a book; Orderly clears it and is not scored either, being order-book infrastructure that other front ends route into rather than a venue a trader opens an account on. Four venues that were scored here until this edition no longer are: edgeX, ApeX Omni and Paradex document three asset classes apiece, dYdX one. Their documentation is still read, and each is named under the table of any page where a reader would otherwise wonder where it went; none of them carries a place or a mark anywhere on this site.

## Where each column was read or measured

- **Read.** What a venue lists, when it says the market is open, what it charges and who is able to move
  the collateral: its own documentation and trading terms, carrying the date they were read.
- **Measured.** Depth: ten-minute readings of the public BTC perpetual book at 28 venues, counting
  resting orders on both sides inside 10 basis points of the mid price.
- **Taken from a publisher.** Open interest and 24-hour volume are each venue's own figure as published
  by CoinMarketCap on 22 September 2026; audit records come from CertiK Skynet.

## The desk that wrote this sheet is paid to publish

A rubric is worth exactly as much as the reader knows about the desk that wrote it. Publication on this
domain is commissioned, and a venue scored by the weights above may be the party that paid. The weights
were settled on 23 September 2026 and the sheet recomputed once against them; nothing was re-weighted
afterwards and no total on this site has ever been moved by hand. The rest of the arrangement is set
out on [about this desk](/about).

## Two venues, one number

Two venues that round to the same total hold the place between them, marked with an equals sign. Inside a shared place the names run alphabetically, which says nothing about
either of them.

## When the sheet is recomputed

The sheet is reviewed once a quarter, and out of turn whenever a venue changes its listings, its hours
or its rates — a documented change is read on the day the desk sees it rather than held for the next
quarter. Should the method itself change, it takes a new number and a new date at the head of this
page, and every older table is recomputed on the new version before it is published — two methods are
never left running side by side.

Weights: What one account reaches 32 · Cost of getting in and out 28 · Custody of the margin 18 · How the venue has held up 12 · The book behind the quote 10, on a scale of 100. Version 1.1, fixed 23 September 2026 and unchanged since.

— Cross-Asset Desk Research. Edition of 24 September 2026.

This desk is paid to publish, and a venue covered here may be the party that paid.