The Perp DEX Liquidation Fairness Index 2027
A liquidation engine protects a market by taking control away from a trader at the worst possible moment. The DN index measures whether that transfer is predictable, proportionate and auditable, or whether hidden penalties, stale prices and unnecessary position closure turn risk management into value extraction.
What Matters
Liquidation is an execution event
The displayed liquidation price is only an estimate. Fairness depends on the trigger source, delay, close price, position fraction, penalty, backstop and remaining account value.
Restore health with minimum destruction
A fair engine should close only what is required to restore solvency, unless speed or liquidity makes partial liquidation unsafe.
Documentation is not performance
Protocol rules explain intended behavior. Only timestamped liquidation records can reveal actual trigger error, value loss and stress-period consistency.
Why Liquidation Rankings Are Usually Misleading
Most venue comparisons treat liquidation as a single number: the price shown beside a position. That number hides the mechanism that determines the final loss. Two platforms can display similar estimates yet produce very different outcomes because one reduces a position incrementally while another transfers or closes it more aggressively.
A credible fairness test separates seven questions:
- Which price triggers liquidation: oracle, mark, index or executable market price?
- How far did the actual trigger deviate from the pre-event estimate?
- Was only the necessary portion closed?
- At what price was the position transferred or executed?
- What penalty, fee and adverse price impact did the trader bear?
- What happened when the insurance or liquidity backstop was insufficient?
- Can an independent observer reconstruct the event?
Fairness does not mean preventing loss. A leveraged trader who falls below maintenance requirements must expect forced risk reduction. Fairness means the mechanism is disclosed, reproducible, proportionate and no more destructive than required to protect market solvency.
How Current Perp DEX Designs Differ
| Venue | Disclosed trigger and mechanism | Position treatment | Backstop | DN evidence status |
|---|---|---|---|---|
| Hyperliquid | Liquidation occurs when account equity falls below maintenance margin. Maintenance requirements vary with maximum leverage and margin tiers. | Its documentation describes liquidation through the protocol's liquidation system and distinguishes cross, isolated and portfolio-margin logic. | Protocol liquidation and backstop mechanisms, with onchain event visibility. | Mechanism documented. A matched event dataset is still required before awarding a performance grade. |
| dYdX | Liquidation is based on subaccount maintenance requirements and protocol liquidation logic. | A subaccount position may be partially or fully closed. | Profits or losses from liquidation are taken by the insurance fund, subject to available capacity. | Mechanism documented. Stress execution and fund-capacity outcomes require observation. |
| GMX | Liquidation depends on remaining collateral after PnL, fees and price impact, with documented minimum collateral thresholds. | Position closure uses GMX's oracle-priced liquidity-pool architecture rather than a conventional central limit order book. | Market pools and ADL rules form part of the risk architecture. | Mechanism and fee inputs documented. Actual event cost must be reconstructed transaction by transaction. |
| Paradex | An account enters liquidation when its margin ratio exceeds 100% during a health check. | Partial liquidation reduces all positions by a calculated share, in 20% increments, targeting a margin ratio below 90%. | A liquidation penalty is paid to the insurance fund. | Strong proportionality disclosure. Timing, execution price and stress consistency still require testing. |
This table compares current official disclosures. It is not a claim that one venue delivered superior live outcomes. Operational status and specifications can change.
The DN Liquidation Fairness Index
DN-LFI scores a venue from 0 to 100 using matched liquidation observations. Scores apply to a specific market, margin mode, observation window and protocol version. A venue does not receive a numerical ranking from documentation alone.
| Dimension | Weight | Measurement | Fair outcome |
|---|---|---|---|
| Trigger integrity | 20 | Difference between disclosed threshold, observed trigger and valid reference price | No premature trigger and consistent price-source use |
| Liquidation proportionality | 20 | Position percentage closed relative to the minimum needed to restore health | Minimum economically safe reduction |
| Value retention | 20 | Equity remaining after penalty, slippage, fees and unnecessary closure | Maximum solvent equity preserved |
| Stress execution | 15 | Latency, completion, price deviation and score retention during volatility | Limited degradation under load |
| Backstop fairness | 10 | Insurance use, ADL rules, loss socialization and priority treatment | Predictable losses without arbitrary transfer to unrelated traders |
| Transparency | 10 | Public formulas, parameters, events, receipts and version history | Independent reconstruction is possible |
| Recovery and recourse | 5 | Error reporting, post-event explanation and remedy process | Documented escalation for incorrect execution |
Hard gates
- Measurable: no grade above C without event-level liquidation data.
- Proportionate: no grade above B if the mechanism routinely closes materially more than required without a disclosed safety reason.
- Stress-tested: no A grade without observations during a defined high-volatility window.
DN Liquidation Fairness Calculator
Enter one observed liquidation event. Use verifiable transaction, account and market data rather than estimates where possible.
Grade F
Estimated avoidable equity loss: $0
The Metric That Matters: Avoidable Equity Loss
Raw liquidation loss mixes two different things. The first is unavoidable: the position moved against the trader and breached maintenance requirements. The second is mechanism loss: excess position closure, penalty, slippage, latency or an inconsistent trigger removed more equity than required.
DN defines Avoidable Equity Loss as the portion of post-trigger loss attributable to the mechanism beyond a disclosed minimum-risk-reduction baseline. It is not always observable from public data, but it is the closest measure of whether a liquidation engine protected the market efficiently or extracted unnecessary value.
For example, a partial liquidation can look trader-friendly but still be unfair if it executes late at a poor price. A full liquidation can be economically defensible in a thin market if incremental reductions would expose the insurance fund to rapidly growing loss. Context matters, which is why every observation must retain market depth, volatility, oracle state and system-load data.
DN Matched-Test Methodology
- Select only venues passing the DN Operational Status Gate.
- Fix the market, collateral type, position direction, leverage band and margin mode.
- Capture the displayed liquidation estimate, maintenance formula and protocol version before the event.
- Record reference prices from the venue's declared trigger source and independent market sources.
- Measure trigger time, transfer or fill price, percentage closed, penalty, fees, equity retained and backstop action.
- Tag ordinary and stress periods separately.
- Publish sample size, exclusions, missing fields and confidence level.
A minimum publishable venue grade should contain 30 valid observations across at least three markets, with at least five stress-window events. Institutional confidence should require 100 or more observations and repeat testing after material risk-engine changes.
DN Alpha Thesis: Liquidation Quality Will Become a Capital Cost
As perpetual markets compete for professional flow, liquidation fairness will stop being a support-page detail. Traders, vaults and autonomous agents will price it into venue selection. A platform that preserves more equity during forced deleveraging can support higher sustainable turnover even if its headline fee is not the lowest.
This creates a defensible DN data product: a versioned event database linking liquidation thresholds, actual execution, market depth, oracle state, penalties and recovery outcomes. The resulting index can power public rankings, enterprise feeds, agent routing rules and venue due diligence. The moat is longitudinal evidence, not another static comparison table.
What Would Prove This Thesis Wrong?
The thesis would weaken if liquidation outcomes converged so tightly across venues that mechanism differences became economically immaterial. It would also weaken if professional traders cared only about liquidity and rebates, accepting liquidation loss as irrelevant because they always self-close first.
DN will watch dispersion in retained equity, excess closure, stress slippage and backstop losses. If these remain negligible after matched testing, liquidation fairness should receive less weight than depth and ordinary execution quality.
Primary Sources
Frequently Asked Questions
What is a perp DEX liquidation?
It is forced position reduction or closure after an account falls below the venue's maintenance requirements. The mechanism protects the market from a position whose losses may exceed available collateral.
Is the displayed liquidation price guaranteed?
No. It is an estimate based on current collateral, fees, funding, margin rules and the venue's trigger price. Those inputs can change before execution.
What makes a liquidation fair?
Predictable trigger rules, proportionate position reduction, limited penalty and price loss, transparent event data, consistent stress behavior and a clear backstop process.
Is partial liquidation always better?
No. It usually preserves more trader equity, but incremental liquidation can be unsafe during extreme volatility or thin liquidity. The venue should disclose why more aggressive closure is necessary.
What is ADL?
Auto-deleveraging reduces positions held by other traders when ordinary liquidation and insurance mechanisms cannot absorb a bankrupt account. Its priority and loss-allocation rules affect fairness.
Why can fees move my liquidation price?
Funding, borrowing and trading fees reduce remaining collateral. As net collateral falls, the maintenance threshold can be reached sooner.
How does an oracle affect liquidation?
If the venue uses an oracle-derived mark or index, update speed, source quality and manipulation resistance affect when liquidation triggers.
Can a stop-loss prevent liquidation?
It can reduce the probability, but it is not guaranteed. A stop may trigger late, fail, partially fill or execute beyond the intended price during stress.
What is Avoidable Equity Loss?
DN's estimate of value lost because of excess closure, penalty, delay, price deviation or mechanism failure beyond the minimum loss needed to restore account health.
Does DN-LFI rank these venues today?
No. Version 1.0 defines the framework and compares disclosed mechanisms. Performance grades require a matched event dataset with adequate sample size.
