# How a venue is scored here

Updated: 24 September 2026 — Session Log staff

Every total on this site comes from one set of weights, fixed on 22 September 2026 and written down
before a single venue on this domain had been read against it. Six criteria, each scored out of ten
on its own scale, carry 30, 30, 15, 15, 5 and 5 of a hundred points; the result is put on a range of
0 to 25 and carried to 1 decimal place. A page decides
which venues it shows. It never decides what a criterion is worth. Who is eligible to be shown at all
is decided earlier still, by the register described below, and the register is about the price of a
crossing rather than about any score. The pages on this domain were commissioned and paid for.

## What each criterion is worth

| Criterion | Weight | What it reads | Why it carries that weight | Marks it actually spanned |
|---|---|---|---|---|
| Cost of a round trip | 30 | What the venue takes for crossing the spread on the way into a BTC perpetual and again on the way out, at the rate a new account pays, counted in basis points of the position. | A day trader pays the taker round trip on every position and often more than twenty times a day, so the entry-tier fee is the one number that compounds with activity rather than with balance. | 3.0 to 10.0 over the ten scored, 6.0 to 8.5 over the seven on the register |
| Measured liquidity | 30 | The dollars found sitting within 10 basis points of the BTC mid, bids and asks together, taken as the median of one session of readings off the public book. | The other half of the cost of every entry and exit: resting depth within ten basis points of mid is what a market order actually meets, it is the only criterion we measure ourselves, and one bad fill can cost more than a day of fees, so it is weighted level with price. | 0.0 to 6.9 over the ten scored, but only 3.9 to 6.9 over the seven on the register |
| Custody and settlement | 15 | Ten points, taken from each venue's own documentation — what reaches a chain after a fill, and what a trader can do when nobody at the venue is answering. | An intraday trader leaves collateral on the venue for weeks at a time, so what is recorded on-chain and whether money can leave without the operator is what a stalled session actually costs. | 5.0 to 10.0 over the ten scored, and the same 5.0 to 10.0 over the seven on the register |
| Security record | 15 | Ten points for what has been audited, what a bug is worth to the venue, and what went missing in the last twenty-four months. | Audits, a published bug bounty and a clean twenty-four-month record are the only public evidence a venue survives the day it breaks, which is the same failure custody describes seen from the other side, so it carries the same weight. | 4.0 to 10.0 over the ten scored, 7.0 to 10.0 over the seven on the register |
| Market coverage | 5 | How long the perpetual market list runs, and how many kinds of thing other than crypto a trader can put a session into. | Intraday volume lives in a handful of markets, so a venue with five hundred listings is not ten times more useful than one with fifty; breadth is scored but kept small. | 5.2 to 10.0 over the ten scored, and the same 5.2 to 10.0 over the seven on the register |
| Access | 5 | Five points for what stands between a wallet and a first order, doubled onto the same ten-point scale. | Whether a wallet signature is enough to start, and which jurisdictions are excluded, decides who can use the venue at all but changes nothing about how a session trades. | 8.0 to 10.0 over the ten scored, and the same 8.0 to 10.0 over the seven on the register |


That last column is there because a weight on its own is only half of a sequence. A criterion can
only push one venue past another as far as their two marks are apart, so thirty points resting on
marks three apart move a table less than fifteen points resting on marks five apart. It shows on
this domain. Across the seven venues on the register, custody and settlement carries fifteen
points and spans five marks, while measured liquidity carries twice the weight and spans three:
the lighter criterion decides more of that order than the heavier one. Both spreads are in the
column above, taken across the seven venues the standings list. The weights are not being re-cut —
changing a weight after the marks are in is choosing the answer instead of reading it — but a
reader who meets a similar order on somebody else's page deserves to see which criteria here could
have separated these venues and which could not.

Sixty of the hundred points sit in the first two rows, and that is the shape of every table on
this site. Extended scores 8.5 for cost of a round trip, 5.8 for measured liquidity, 7.0 for custody and settlement, 10.0 for security record, 9.6 for market coverage and 8.0 for access; carried in the proportions above and put on
the site range, that is 19.3 out of 25. Paradex charges a
taker nothing at all, takes the full ten on cost and still finishes below six venues that pay,
because the book we read behind it was the thinnest on the desk.

## Cost of a round trip

Every venue opens the day on ten points. A round trip that takes two basis points or less keeps the lot; past that, half a point comes off for each further basis point the crossing costs, so eleven basis points a side — twenty-two on the round trip — spends the whole ten and leaves nothing. A schedule with no taker fee on it never starts spending.

Worked example. Extended charges 0.025% to a taker
([Extended fee page](https://docs.extended.exchange/extended-resources/trading/trading-fees-and-rebates), checked 18 September 2026): five basis points there and back, three
above the ceiling, half a point each, and 8.5 of ten survives. Aevo charges 0.080%
([Aevo fee page](https://www.aevo.xyz/docs/aevo-products/aevo-exchange/fees/perpetuals-fees), checked 18 September 2026) — sixteen basis points on the round trip, which
leaves 3.0. Lighter and Paradex publish no taker fee on the account a new trader opens, so both
take the full ten and this criterion stops telling them apart.

Which is why the register is decided before the scoring and not by it. A venue is ranked on this
site when its entry tier takes five basis points or less to cross, at a price not bought with added delay, because a session of forty
crossings carries that rate forty times. Aevo is outside the rule on price. Lighter and Paradex
are outside it although they publish nothing, because each buys the zero with added latency on
the order paying it: Lighter's fee page gives "Taker latency: 300ms" for the free Standard
account, and Paradex documents a 300 ms speed bump on retail orders and none on the paid route. A
free trade that arrives late is not free to a trader working the same day. The formula still
scores all 10; three of them carry a total and appear in no ranking.

## Measured liquidity

The ladder starts at nothing: a book holding ten thousand dollars inside the band is worth zero here. Each tenfold step above that adds two points — a hundred thousand, two; a million, four; ten million, six — and the climb stops at ten. The deepest book among the ten scored here, close to thirty million dollars at the mid, is still well short of that ceiling.

Worked example. The Paradex book held a median $245.00 within ten basis points of
the mid across 126 readings, which is 0.0 of ten. The edgeX book held
$29.3 million across 127 successful readings, roughly 120,000 times as much, and takes 6.9. Lighter and Paradex tie on cost, and nearly all of the
5.1
points between their totals comes out of this one column.

## Custody and settlement

Ten points, given like this:

- every fill reaches a chain, batched into a rollup or carried in a validity proof — 4 points
- nothing but a position snapshot reaches a chain, and only now and then — 2 points
- the documentation describes an exit that works while the operator does not — 3 points
- the contracts holding the collateral have been through an audit — 2 points
- sign-in is a signature rather than a registration — 1 point

The first two lines are alternatives rather than additions: a venue is read as recording every
trade or as recording positions now and then, never as both.

Worked example. ApeX Omni takes all ten. Matched trades are proven to the rollup contracts that
hold the deposits ([ApeX Omni documentation](https://docs.zk.link/architecture/transactionflow), checked 18 September 2026), a forced
withdrawal to layer 1 without the operator is documented, those contracts are audited and a
wallet is enough to sign in. EVEDEX takes 5.0 on the same rubric
(EVEDEX trading terms and documentation, 18 September 2026): position data reaches Arbitrum only after a batch of
changes builds up, never trade by trade, which is the two-point half of the first line, and
no way to withdraw past the operator appears in the documentation, so the three points for one go
unclaimed. The audited deposit contract and the wallet signature account for the rest.

## Security record

Ten points, given like this:

- an outside audit of the code is worth two points, and a second one two more — 4 points
- a bug bounty open to anyone who finds something — 3 points
- twenty-four months on the public record with no user money lost — 3 points

Worked example. Aster has seven audit reports behind it and edgeX six; both stop at four points,
because counting stops at the second audit. Aster adds a public bug bounty and a clean
twenty-four months and reaches ten. edgeX has no bounty listed anywhere we looked and stops at
7.0. Aevo takes 4.0 — five audits, no bounty, and a loss inside the window.

## Market coverage

Six points at the outside for the length of the market list, on a curve that flattens early: five hundred perpetuals collects all six, while a board of fifty still collects nearly four. Anything listed that is not crypto is worth a point a class, and the fourth is the last one counted.

Worked example. dYdX documents 78 perpetual markets and one asset class beyond
crypto, which comes to 5.2 of ten. Aster documents 578 markets across five
classes, reaches both ceilings and takes ten. The curve is steep at the bottom and flat at the
top on purpose: a dozen markets against a hundred matters to an intraday trader, three hundred
against six hundred does not.

## Access

Five points, doubled onto the same ten-point scale, given like this:

- no identity papers are asked for before an order can be sent — 2 points
- the venue opens on a wallet signature alone — 1 point
- the terms close the door on sanctioned parties and a short roll of countries besides — 2 points
- the terms close it on a wider list than that — 1 point

The last two lines are alternatives, not additions.

Worked example. ApeX Omni is the only venue on the desk at ten: its documentation asks for no
forms, a wallet or an email opens the account, and its terms stop at sanctioned persons and a
single national exclusion. Every other scored venue shuts out a longer list of countries and
lands on 8.0, so in practice this criterion separates one venue from the other nine and nothing
else.

## What counts as a loss of user funds

A loss means user money gone — taken in an exploit, drained out of a venue-run vault by manipulation, or left unrecovered after an outage the venue does not make good in full. A liquidation that simply follows the market is not a loss here.

Three venues on the desk lose those three points under that wording. Four more had an incident
inside the same window and keep them, because the money was returned in full or because what
moved was the market rather than the venue.

## Which venues are scored, and which are only listed

Venues running a central limit order book, whether that book sits off-chain with settlement on a chain or on a chain of its own. Venues priced by an oracle or by a pool appear with their facts and no score, because the liquidity criterion reads resting orders they never have.

That leaves 10 venues with a total on this domain. The rest appear on the
[venue list](/crypto-day-trading-platform-list) with their fees, markets, sign-in rules and
restrictions printed in full and no score beside them. That is a statement about the formula, not
about the venue.

A total is still not a place. The register — the venues this log actually ranks — is narrower
than the scored field: seven, the ones whose entry tier takes five basis points or less to cross, at a price not bought with added delay.
The other three carry a total, appear on the front page and the venue list with every figure the
desk holds for them, and are ranked on nothing. Where a head-to-head page compares one of them,
it says on the page that the venue is off the register and why.

## Where each figure comes from, and when

Five of the six criteria are read rather than measured. Fee schedules, market counts, asset
classes, sign-in rules, restricted jurisdictions and audit records come from each venue's own
documentation, terms and margin schedule, all of it read on 18 September 2026, and each figure
keeps the day it was read.

One criterion is not read anywhere. Every BTC perpetual book on the desk was polled
six times an hour until the log closed on 22 September 2026, and what enters the formula is the median of
the successful readings — not the deepest, not the last, and never a number a venue published
about itself.

Open interest and 24-hour volume appear on other pages of this site and in no criterion at all.
Both are the venue's own number, lifted from CoinMarketCap's derivatives ranking on 22 September 2026.

## Equal totals, and what sets the order of a page

Two venues on the same total after rounding are shown at the same place and printed as tied. Inside a tie the sequence is alphabetical and carries no meaning.

A page built around one measurement is sorted by that measurement; everything else is sorted by
the total:

- The total each venue carries, highest first — Home, Standings, Score and ladder, Venue list
- Deepest book first, by the dollars found resting inside ten basis points of the BTC mid — Alternatives, Order books
- Cheapest crossing first, by what one $10,000 round trip takes at the entry tier — Fees, Scalping

Neither the weights nor the totals shift from one page to the next. A venue's score is the same
number everywhere on this site, and a page sorted by a single measurement still prints that score
in its table.

## What the total cannot see

**Discounts.** The cost column is the entry tier, what a new account pays by default. Volume
tiers, maker rebates, referral codes and cashback schemes all move the real number, and each of
them turns on a trader's own record rather than on the published schedule.

**Conduct.** Custody and security are read from documents, audit reports and the public record of
past incidents. How a venue behaves in the hour it breaks, and whether a promised refund arrives,
is in no column above.

**Every market that is not BTC.** Depth, spread and price impact were measured on one contract.
A deep BTC book can sit beside thin ones on the same venue, and the formula would never know.

**The hour it was read.** The window behind the liquidity column is one session, closing on
22 September 2026, and it was a quiet one. A book read in a crash is a
different book.

## When the weights change, and how a figure is corrected

These are the weights as fixed on 22 September 2026, version 1.0. The desk is
read again at the start of each month, and sooner when a venue republishes its fee schedule or a book
moves far enough to change a place. A new set of weights takes a new number and a new date at the
top of this entry and every venue is rescored under it; totals from two versions are never
printed in one table.

A wrong figure is corrected where it is stored, so every total leaning on it moves at the same
time and the session date moves with the entry. Corrections go to
editorial@coinlen.com — name the entry and point at the document that says otherwise.

Corrections: editorial@coinlen.com.

Session Log staff, 24 September 2026

This site is paid placement.