Extended: perpetual dexs rating breakdown
StarkEx-based, 2 chains
A clean, validity-proof-based venue in the lineage that dYdX V3 came from: matching happens off-chain, correctness is proven on-chain, and withdrawals do not depend on the operator being cooperative. $124m of collateral and $1.57m of thirty-day fees put it firmly in the working middle of this category.
A clean, validity-proof-based venue in the lineage that dYdX V3 came from: matching happens off-chain, correctness is proven on-chain, and withdrawals do not depend on the operator being cooperative. $124m of collateral and $1.57m of thirty-day fees put it firmly in the working middle of this category.
How the score is built
Each criterion is scored 0–10 and weighted. The median column is the middle score across every entry in this ranking, so a row reads as a position rather than a number.
| Criterion | What we measured | Weight | Score | Category median | Rank | Weighted gap |
|---|---|---|---|---|---|---|
| What happens when the rules cost the house money | — | 30% | 6 | 7 | 6 of 8 | -0.30 |
| Depth and where the flow actually is | $1.57m 30-day protocol fees · sourced · 2026-08-27 · source$124m value locked · sourced · 2026-08-27 · source | 25% | 4 | 5 | 6 of 8 | -0.25 |
| What a position really costs | — | 22% | 7 | 6.5 | 3 of 8 | +0.11 |
| Who holds the collateral | — | 15% | 7 | 7 | 4 of 8 | 0.00 |
| Markets and access | — | 8% | 5 | 6 | 6 of 8 | -0.08 |
Measured 27 August 2026 · weights and method · decided by what happens when the rules cost the house money, worth -0.30 points against the median
What happens when the rules cost the house money: 6/10
Every intervention on record: forced settlements, delistings, socialised losses, auto-deleveraging events and oracle changes, with dates and amounts. Hyperliquid settled an entire manipulated market at a chosen price and delisted it by validator vote in March 2025, then reimbursed users from a foundation. That is a specific, documented answer to “who decides”, and it is worth more than any decentralisation claim on a landing page.
Scored 6 of 10 against a category median of 7, which places it 6th of 8 among perpetual dexs on this criterion. At a 30% weight that is 0.30 points below the median contribution of the weighted total. The best score in the category is 8, the worst 5.
Depth and where the flow actually is: 4/10
Collateral and vault value locked at a stated timestamp, alongside thirty days of protocol fees as the closest public proxy for real flow. Depth on a perpetual venue is not a comfort — it is what decides whether your stop fills near your stop.
Scored 4 of 10 against a category median of 5, which places it 6th of 8 among perpetual dexs on this criterion. At a 25% weight that is 0.25 points below the median contribution of the weighted total. The best score in the category is 10, the worst 4.
What a position really costs: 7/10
Published taker and maker fees, plus the funding mechanism as documented — and then our own division: thirty days of protocol fees, annualised, per dollar of collateral held. That last number is the intensity of extraction at each venue, and no fee page shows it.
Scored 7 of 10 against a category median of 6.5, which places it 3rd of 8 among perpetual dexs on this criterion. At a 22% weight that is 0.11 points above the median contribution of the weighted total. The best score in the category is 9, the worst 5.
Who holds the collateral: 7/10
Whether margin sits in a contract you can verify or on a company balance sheet, who runs the sequencer or validator set, and how upgrades and market listings are authorised. A venue where a small set of operators can pause, list or settle is a custodian with extra steps.
Scored 7 of 10 against a category median of 7, which places it 4th of 8 among perpetual dexs on this criterion. At a 15% weight that is exactly level with the median of the weighted total. The best score in the category is 9, the worst 6.
Markets and access: 5/10
Markets listed, chains supported, and the practical route in and out — a venue that only accepts one collateral asset on one chain is narrower than its market count suggests.
Scored 5 of 10 against a category median of 6, which places it 6th of 8 among perpetual dexs on this criterion. At a 8% weight that is 0.08 points below the median contribution of the weighted total. The best score in the category is 9, the worst 4.
Its nearest neighbours in this ranking
Questions about this score
What is the difference between validity proofs and trusting the operator?
+
A validity proof lets anyone verify that the state transition — including your withdrawal — followed the rules, so exiting does not require the operator's cooperation or goodwill. That closes the failure mode where a venue simply stops processing withdrawals. It does not cover transaction ordering, which the operator still controls, so front-running and sequencing remain trust assumptions.
Has Extended been tested in a crash?
+
Not at size, and we say so rather than inferring safety from the absence of incidents. A perpetuals venue is defined by its behaviour during a violent liquidation cascade, and Extended has not yet had one at meaningful scale. Design quality is evidence; a stress record is proof, and it does not have one yet.