How a venue gets its score out of sixteen

The share each criterion takes out of a hundred

The scored list runs to eight venues. Each of them keeps a public central limit order book on a BTC perpetual contract, which is the condition for being scored at all: the queue has to be readable from outside, without an account and without sending an order. A venue whose price arrives from an oracle, and whose fill is drawn out of a shared pool, is left off the scored list — the criterion carrying the largest share of the total has nothing to read there, and its mark would have to be borrowed from a figure somebody else published. That is a limit on what this instrument measures, not a judgement on those venues.

The shares below add to a hundred and are identical on every page of this site.

The last column is the one that decides how much a weight is worth in practice, and it is printed here because a weight on its own says nothing. A criterion moves a table only as far as the marks underneath it are apart, so thirty points resting on marks that sit within a mark of each other shift a sequence less than fifteen points resting on marks five apart. Across the eight venues on the sheet further down, the order is close to what the weights imply. Across the six on the futures table it is not: depth carries thirty points and separates those six by 1.1 marks, settlement and withdrawal carries fifteen and separates them by 5.0, and getting started separates them by nothing at all, because all six hold the same mark for it. None of the weights has been adjusted to correct for that, and none will be — moving a weight once the marks are in is choosing an answer rather than reading one — but a reader holding this table against another site's is owed the figure.

What a mark out of ten means

Each criterion is marked out of ten before any weight touches it, so a mark carries the same meaning whichever criterion earned it. Where the rule is a list of points, the lines add and the total stops at ten. The number in each heading is the weight that mark carries in the total.

Order-book depth, 30 points of the hundred

The column is scored on the width of the figure rather than the figure itself. Ten thousand dollars standing inside the band prints nothing, and a thinner book prints nothing either; above that the mark rises two points for every tenfold, smoothly rather than in steps, so a hundred thousand reads two, a million four and ten million six. Ten is where the column stops printing. Reaching it would take a billion dollars standing inside the band, which is why the deepest book read on this site still prints just under seven.

A futures venue publishes its fee schedule and hides its book, so depth is the cost a trader cannot look up. The gap between the deepest and the thinnest BTC book in this capture cycle is $29.2 million against $728 — far larger than the gap between the cheapest and dearest fee tier. It is also the only number on this site that is measured every ten minutes rather than accepted from a venue.

Round-trip cost, 25 points of the hundred

Ten stands while the round trip clears in two basis points or less. Past two the mark counts down half a point for every further basis point, which prints a 0.045% taker rate — nine basis points across the two fills — at six and a half, and empties the column exactly at twenty-two. Dearer than that and the cell reads zero; there is nothing below zero to print.

The taker fee is paid on the way in, on the way out and again on every resize of a perpetual position, so it compounds with turnover in a way nothing else here does. Second weight rather than first because it is published, comparable and easy for a reader to check without us.

Loss and audit record, 15 points of the hundred

• an independent security audit, two points each, counted up to two audits — 4 points
• a public bug bounty programme with a published ceiling — 3 points
• no loss of user money in the last 24 months — 3 points

A loss here means user money taken in an exploit, drained out of a venue-run vault by somebody moving the market, or stranded because the venue itself stopped working, and never made whole afterwards. A position closed out because the market went against it is not a loss under that rule.

Margin sits at the venue between trades, so what matters is the record: how many independent audits, whether there is a public bounty, and whether user funds were lost in the last 24 months. Kept below the two execution criteria because it is a record of the past, not a price paid on every trade.

Settlement and withdrawal, 15 points of the hundred

• every trade written or proven on a public chain, in a batch or in a validity proof — 4 points
• only a position snapshot written on-chain from time to time — 2 points
• a documented withdrawal that works when the operator does not — 3 points
• the contract holding deposits has been audited — 2 points
• a wallet signature is the whole account, with nothing to register — 1 point

The first two lines are alternatives rather than a sum: a venue is marked on the strongest settlement its documentation describes, not on both. The three lines under them add on top.

Whether every trade is written or proven on a public chain, or only a position snapshot is, and whether a documented withdrawal works when the operator does not. Equal to the security record because the two answer the same question from different ends: what happens to the collateral if the venue stops.

Contracts listed, 10 points of the hundred

Six points ride on the length of the board and four on its breadth. The length is compressed rather than added up: five hundred perpetual contracts fills all six, a hundred prints about four and a half, and every contract after that is worth less than the one before it, with nothing above five hundred adding anything at all. Breadth is counted straight — one point per asset class outside crypto, and the fourth is the last that prints.

How many perpetual contracts a venue lists and how many asset classes beyond crypto share one margin balance. Real but secondary: most futures traders use a handful of markets, and a long list does not improve the fill on BTC.

Getting started, 5 points of the hundred

• no identity documents asked for before a first trade — 2 points
• a wallet signature is enough to sign in — 1 point
• the refused list stops at sanctioned jurisdictions and one large market — 2 points
• the refused list runs wider than that — 1 point

The last two lines are alternatives, so the four are worth five points between them, and the mark is twice whatever they add to. That is the only criterion here where a raw count is stretched to fit the ten-point scale.

Wallet sign-in, what documents are asked for before a first trade, and how wide the restricted list is. Lowest weight because it is a one-time hurdle, while every other criterion is paid again on every position.

One venue, worked through

A rule described in words is worth checking against one row a reader can rebuild, so edgeX is taken through all six criteria here, on the documents read on 18 September 2026 and the book read in the capture cycle below.

Depth. Its BTC book held a median $29.2 million resting within ten basis points of the mid across 251 readings taken here. That is 2,917 times the ten thousand dollars the scale starts from, three and a half tenfold steps above it, and two points a step gives 6.9.

Cost. A taker pays 0.045% going in and the same coming out, so the round trip costs nine basis points of the position. Seven of those nine sit beyond the two that cost nothing, at half a point each, which takes 3.5 off ten and leaves 6.5.

Loss and audit record. Six audits stand on the public record — RigSec, SlowMist twice, Halborn, Zellic and Spearbit — and the rule counts the first two of them at two points each, so four. No bug bounty page is published, so that line is empty. No user money was lost at the venue in the twenty-four months to 18 September 2026, so three more. Seven of ten.

Settlement and withdrawal. Trades and balances settle on the venue's own rollup, worth four; a withdrawal that works while the operator does not is documented, worth three; the contract holding deposits has been audited, worth two; an account is a wallet signature with nothing to register, worth one. Ten of ten.

Contracts listed. 172 perpetual contracts earns just under five of the six points the length of a list can carry, and three asset classes beyond crypto add one point each. Eight of ten.

Getting started. No identity documents are asked for before a first trade, worth two; a wallet signature signs in, worth one; the refused list runs wider than sanctioned jurisdictions and one large market, worth one. Four points of the five available, doubled to eight.

The total. Multiply each mark by its share of the hundred — depth 6.9 by thirty, cost 6.5 by twenty-five, the audit record 7.0 and settlement 10.0 by fifteen each, contracts 8.0 by ten, getting started 8.0 by five — then add those six results and divide by a hundred, which comes to 7.45 on the ten-point scale. Multiplied by 1.6 that is 11.9 out of 16, the number printed beside edgeX wherever it appears here.

From six marks to a number out of sixteen

The weighted average lands on a scale running from zero to 16. Sixteen rather than ten or a hundred so that no one reads a score as a percentage of anything: a total of 11.9 out of 16 is six marks weighted and stretched onto this site's scale, not a proportion of any standard a venue could pass or fail.

One decimal place is published, for a total and for each mark behind it. The arithmetic itself runs on unrounded marks, so adding the six figures in the table below can leave a reader a hundredth off what that table prints. The published total settles the disagreement, because it is the figure every sequence on this site is sorted on.

Nothing about a tie is decided by hand. Totals that come out level after rounding take one place between them, and the page states that they are level instead of putting either one above. Where a page sorts on one column, two identical readings fall back on the overall score, and two identical fee schedules hold one position between them, in alphabetical order.

Which column sets the sequence is a page's own choice — the overall score on most of them, measured BTC depth on the bitcoin page, the cost of a round trip on the fees page, an outside vendor's security mark on the security-rating page — and each names its column in the line above its table. A page also decides which venues it lists at all. Neither decision moves a weight, and no page publishes a weight of its own.