Perp DEX leverage terms: room to liquidation and what is left
Adverse price move that liquidates a position, read from each venue's own published parameters and checked against real liquidations. The selector sets the leverage: the default puts each venue at its own maximum, which is not a like for like reading, and the fixed settings put every venue on the same leverage. No row is ranked, because more room is not the same as better terms.
Read this carefully
Parameters each venue publishes, converted to one model and checked against liquidations that happened. Gains: threshold, max leverage and the pair fee by eth_call on the Arbitrum diamond, checked against 11 liquidations. Papertrade: the instrument tuple on /state/trading, checked as a ceiling on 384 adverse closes. dYdX: both margin fractions on the indexer. Hyperliquid publishes no maintenance requirement, so that row is derived from its documented halving rule.
Four figure leverage is the headline every perp venue leads with, and it says almost nothing on its own. Leverage is a denominator. What a trader needs is the numerator: how far the price can move against the position before the venue closes it, and how much of the collateral comes back when it does.
Each venue reduces to one parameter, the maintenance requirement it holds back as a share of the position's notional, plus whatever fees come out of the same collateral. What separates them is what sets that requirement. Gains ties it to the collateral, so its room falls close to proportionally with leverage. Papertrade fixes it in price at about five basis points, so it costs almost nothing at 10x and roughly half the margin at 1000x. The ranking inverts depending on which leverage you ask about, which is why the leverage selector is part of the measurement rather than a convenience.
Three of the four venues publish every input. Hyperliquid publishes its maximum leverage but not its maintenance requirement, which its documentation fixes at half the initial requirement, so that row alone is derived.
Methodology
For each venue the bench computes the adverse price move, in percent, that liquidates a position opened at the selected leverage on the market shown, the same figure in the unit venues quote among themselves, the venue's maximum leverage, and the share of collateral returned on liquidation. Everything is re-read every 5 minutes by the perp-cohort-stats harness. One model covers all four. A position at leverage L posts 1/L of notional as margin and is liquidated when its equity reaches the maintenance requirement m, also a share of notional. Any fee charged on the way in and out comes out of that same collateral, so the price has that much less distance to travel: room = 1/L - m - 2 x fee collateral lost = room x L dYdX publishes m directly. Hyperliquid's is half its initial requirement at maximum leverage. Gains publishes a liquidation threshold as a share of collateral, which makes m the complement divided by the leverage, and it charges a position size fee on both legs. Papertrade publishes a bust buffer already expressed as a share of price, and charges nothing on either leg. The fee term is not a modelling choice. Gains' threshold is 80 percent of collateral on the x500 pairs, which without fees would put the room at 0.1600 percent of price. Eleven real liquidations fired at a median 59.4 percent of collateral, 0.1189 percent of price, and two legs of the 0.02 percent fee at 500x are exactly the 20 point gap.
Results: 4 venues measured, 0 ranked by Room to liquidation (24h avg)
| № | Venues | Room to liq | Collateral lost | Max leverage | Collateral returned | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Measured, not ranked: sample below the ranking floor | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| not ranked | Hyperliquid | 1.25% | 50.0% | 40.0x | 50.0% | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| not ranked | dYdX v4 | 0.80% | 40.0% | 50.0x | 60.0% | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| not ranked | Gains | 0.12% | 60.0% | 500.0x | 0% | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| not ranked | Papertrade | 0.052% | 52.4% | 1000.0x | 0% | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Frequently asked
What does room to liquidation mean?
It is the percentage the price has to move against a position before the venue closes it, for a position opened at the selected leverage with no margin added beyond the minimum. It is measured at the moment the position opens and excludes funding and borrowing, which push the real trigger closer every hour the position is held.
Why does the order change when I change the leverage?
Because venues set their maintenance requirement against different quantities. Gains ties it to collateral, so its room falls close to proportionally with leverage. Papertrade fixes it in price at about five basis points, so it barely matters at 10x and consumes roughly half the margin at 1000x. At low leverage Papertrade leaves the most room of the four; at its own maximum it leaves the least.
Why is nothing ranked?
Because more room is not the same as better terms. Room comes from holding back less maintenance margin, and the venue that holds back least is the one that keeps everything when it fires. A single ranking would reward that, whichever venue came out on top. The rows publish their figures and the reader picks the leverage that matches the question.
Why does Gains show less room than its own threshold implies?
Because the fee comes out of the same collateral the threshold is measured against. Its contract liquidates at 80 percent of collateral, which alone would be 0.16 percent of price at 500x, but the position pays 0.02 percent of notional on the way in and again on the way out, and at 500x those two legs are 20 points. Eleven real liquidations fired at a median 59.4 percent of collateral, which is 0.1189 percent of price, against the 0.12 percent this model predicts.
Does any venue give collateral back on a liquidation?
The two margin fraction venues do: dYdX returns 60 percent and Hyperliquid 50 percent at their own maxima. Neither high leverage venue does. Gains returned nothing on all 27 liquidations measured across every pair from 7.5x to 500x, and Papertrade's documentation calls crossing the bust price a hard bust that forfeits the full isolated margin.
How is the Papertrade row checked when its contract is unverified?
Against external truth rather than an ABI. Its price field matches the live Hyperliquid mid under the per instrument factor from its own instrument tuple, its leverage field caps at exactly the 1000 its documentation states, and under the resulting reading the predicted bust distance is never crossed on 384 adverse closes, the closest reaching 92 percent of it. A ceiling respected 384 times is a measurement, not a decode that happened to look right.
Why only BTC and ETH?
They are the only two markets all four venues list. Papertrade lists exactly two instruments; the other three list hundreds, but comparing one venue's BTC terms against another's altcoin terms would not mean anything.
Source code github.com/ChainBench/OpenChainBench/tree/main/harnesses/perp-cohort-stats