Paper: arXiv 2611.26623

Authors: Christoph Schlegel, Akaki Mamageishvili

Abstract

We characterise path-independent auto-deleveraging (ADL) rules as an extended log-exponential family. Immunity to account splitting, merging and reopening further restrict the class. Homogeneity (essentially) restricts the class to a one-parameter family ordered by both fairness and platform risk. Its split-proof members continuously interpolate between proportional division and water-filling. Among these rules, decreasing the parameter reduces risk at the cost of increasing inequality in how accounts are treated. This suggests selecting the parameter through an ex-ante calibration to a risk bound on historical data to choose a maximally fair rule subject to that bound. A counterfactual replay of Hyperliquid’s ADL episode on 10 October 2025 finds that a range of intermediate parameters significantly improves on the observed allocation in both burden dispersion and subsequent risk statistics.

Complexity vs Empirical Score

  • Math Complexity: 8.5/10
  • Empirical Rigor: 6.0/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper presents a highly mathematical axiomatic characterization of ADL rules, demonstrating significant theoretical novelty. While it includes a counterfactual replay, the empirical section is more illustrative than exhaustive, focusing on a single historical event. The blend of advanced theory and practical application to a real-world problem positions it well.

Research Flowchart

  flowchart TD
    A["Research Goal: Characterize Path-Independent ADL Rules & Optimize for Risk/Fairness"] --> B{"Methodology: Extended Log-Exponential Family & Homogeneity Analysis"};
    B --> C{"Immunity Constraints: Account Splitting, Merging, Reopening"};
    C --> D["Computational Process: Derivation of One-Parameter Family & Interpolation"];
    D --> E{"Data/Inputs: Historical ADL Episode (e.g., Hyperliquid 10/10/2025)"};
    E --> F["Computational Process: Ex-Ante Risk Calibration & Counterfactual Replay"];
    F --> G["Key Findings: Trade-off between Risk & Inequality; Improved Allocation via Intermediate Parameters"];

Evidence Card

What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card

Proposes

The paper characterises path‑independent auto‑deleveraging (ADL) rules as an extended log‑exponential family and proposes a one‑parameter menu of split‑proof rules that continuously interpolate between proportional division and water‑filling, with the parameter calibrated ex‑ante to a risk bound on

Source sentence
We characterise path-independent auto-deleveraging (ADL) rules as an extended log-exponential family. Its split-proof members continuously interpolate between proportional division and water-filling.

Assumes 6

  • Homogeneity of accounts is assumed, restricting the rule class to a one‑parameter family. model
    Source sentence
    Homogeneity (essentially) restricts the class to a one-parameter family ordered by both fairness and platform risk.
  • The rule must be split‑proof, merge‑proof and reopening‑proof, implying immunity to account splitting, merging and reopening. model
    Source sentence
    Immunity to account splitting, merging and reopening further restrict the class. Proportional division is the only member that is both split- and merge-proof, and reopening-proofness requires entry-price independence.
  • Historical market data can be used to calibrate the parameter by minimizing dispersion subject to a risk bound. data
    Source sentence
    This suggests selecting the parameter through an ex-ante calibration to a risk bound on historical data to choose a maximally fair rule subject to that bound. This gives a simple one-parameter menu for rule design when homogeneity is acceptable and a platform can use historical data to select a parameter from the menu.
  • Evaluation of allocations is passive, holding residual positions without subsequent trading, collateral transfers, funding, liquidations or price impact. execution
    Source sentence
    The passive evaluation omits subsequent trading, collateral transfers, funding, liquidations and price impact.
Show 2 more
  • Entry‑price dependence is excluded for reopening‑proofness, i.e., the rule cannot use entry prices to improve later treatment. model
    Source sentence
    Proportional division is the only member that is both split- and merge-proof, and reopening-proofness requires entry-price independence.
  • Fees are ignored when closing and immediately reopening a position, preserving position and equity. other
    Source sentence
    Closing and immediately reopening a position resets that price while, ignoring fees, preserving both position and equity.

Data

  • HyperMultiAssetedADL · ConejoCapital crypto-onchain public
    Source sentence
    The underlying data come from ConejoCapital’s HyperMultiAssetedADL (ConejoCapital, 2025a) and HyperReplay (ConejoCapital, 2025b) repositories, which provide ADL fills, market data and clearinghouse snapshots from Hyperliquid’s public archive, with support from Hydromancer.
  • Binance price series · Binance market-prices public
    Source sentence
    We rebuild the three-second mark from its documented for-mula (Hyperliquid, 2026b) using Hyperliquid’s archived order books and trades and Binance prices, pinned to Hyperliquid’s archived one-minute marks and the batch prices.

Recognised sources: Binance

Costs

Unclear - sources disagree The platform classifier and the paper-text extraction give different answers, so no value is asserted.

Platform classifier: Not applicable

Paper text extraction: Not modeled

Source sentence (the authors' words)
Closing and immediately reopening a position resets that price while, ignoring fees, preserving both position and equity. The passive evaluation omits subsequent trading, collateral transfers, funding, liquidations and price impact.

Finds 5

  • Every evaluated parameter θ ≤ −0.5, including water‑filling, improves on the observed Hyperliquid allocation in every fairness and risk statistic. performance
    Source sentence
    7.3 Results Every evaluated θ ≤−0.5, including water-filling, improves on the observed allocation in every statistic in Table 1.
  • Intermediate parameters (e.g., θ = −1) achieve better fairness than water‑filling while retaining lower risk than the observed allocation. performance
    Source sentence
    Intermediate parameters improve fairness further while retaining lower risk than the observed allocation. The observed allocation’s high burden Gini reflects the concentration of reductions into almost complete closures for affected accounts, while unaffected accounts bear no reduction.
  • Water‑filling yields more equal percentage reductions than the observed allocation on all three fairness metrics. performance
    Source sentence
    Water-filling itself has more equal percentage reductions than the observed allocation on all three fairness statistics.
  • Proportional division weakly dominates positive‑parameter rules in both fairness (dispersion of percentage reductions) and risk. theoretical
    Source sentence
    With fairness measured by dispersion of percentage reductions, proportional division weakly dominates positive parameters in both fairness and risk.
  • The observed allocation’s breach share at 24 hours (1.176%) is higher than that of proportional division (2.44% at three days) and higher than the breach share for θ = −1 (0.563%). statistical
    Source sentence
    For example, θ = −1 gives a 24-hour breach share of 0.563%, compared with 1.176% under the observed allocation. At three days, the observed breach share is 3.67%, exceeding even proportional division’s 2.44%.

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 19 related papers). Coverage is partial, so this is neither replication nor refutation.

Code

No repository by the authors found in the paper.

Extracted 2026-11-20 · same card as JSON · A language model chose these items: check the source sentence before relying on one.

Cited References (46)

  1. Auto-Deleveraging ( ADL ) (2026) — link
  2. Risk-Based Auto-Deleveraging — → 2603.15963
  3. CME Rulebook, Rule 802: Protection of Clearing House (2026) — link
  4. Contract loss mechanisms on dYdX Chain (2026) — link
  5. edgeX Perpetuals: Auto-Deleveraging ( ADL ) Mechanism Explained (2026) — link
  6. Liquidation Logic (2026) — link
  7. Auto-deleveraging (2026) — link
  8. Robust price indices (2026) — link
and 38 more
  1. Clearing Rules, Rule 915: Partial Tear-Up (2025) — link
  2. Parametric rationing methods (2006)
  3. Liquidations & LLP ( Insurance Fund ) (2026) — link
  4. Socialized Losses and Last Resort Mechanism (2026) — link
  5. Robinhood Wallet perpetual futures (2026) — link
  6. Bi-Compositional Division Rules (2026) — link
  7. On Dividing an Amount According to Individual Claims or Liabilities (1987) — link
  8. Autodeleveraging: Impossibilities and Optimization — → 2512.01112
  9. Autodeleveraging as Online Learning — → 2602.15182
  10. Recovery of Financial Market Infrastructures (2014)
  11. On the Recovery Tools of a Central Counterparty (2023) — link
  12. Fundamental Questions on Central Counterparties: A Review of the Literature (2021) — link
  13. Central Counterparties in Crisis: International Commodities Clearing House , New Zealand Futures and Options Exchange and the Stephen Francis Affair (2016) — link
  14. A Problem of Rights Arbitration from the Talmud (1982) — link
  15. Game Theoretic Analysis of a Bankruptcy Problem from the Talmud (1985) — link
  16. Axiomatic Cost and Surplus Sharing (2002) — link
  17. Axiomatic and Game-Theoretic Analysis of Bankruptcy and Taxation Problems: A Survey (2003) — link
  18. Consistency and Its Converse: An Introduction (2011) — link
  19. On the Axiomatics of Resource Allocation: Interpreting the Consistency Principle (2012) — link
  20. Axiomatic and Game-Theoretic Analysis of Bankruptcy and Taxation Problems: An Update (2015) — link
  21. How to Divide When There Isn’t Enough: From Aristotle , the Talmud , and Maimonides to the Axiomatics of Resource Allocation (2019) — link
  22. Price Discovery in Bitcoin : The Impact of Unregulated Markets (2020) — link
  23. The BitMEX Insurance Fund (2019) — link
  24. XBTU19 and ETHM19 Auto-Deleveraging Events 2 April 2019 (2019) — link
  25. Financial Stability Risks from Cryptoassets in Emerging Market Economies (2023) — link
  26. The Evolution of Decentralized Exchange: Risks, Benefits, and Oversight (2026) — link
  27. Distributive Justice in Taxation (1988)
  28. Coherent Measures of Risk (1999) — link
  29. Priority Rules and Other Asymmetric Rationing Methods (2000) — link
  30. Asymmetric Rules for Claims Problems without Homogeneity (2006) — link
  31. Coalitional Manipulations in a Bankruptcy Problem (1999) — link
  32. Non-Manipulable Division Rules in Claim Problems and Generalizations (2007) — link
  33. Lorenz Rankings of Rules for the Adjudication of Conflicting Claims (2012) — link
  34. Spectral Measures of Risk: A Coherent Representation of Subjective Risk Aversion (2002) — link
  35. Least $d$ -Majorized Network Flows with Inventory and Statistical Applications (1971) — link
  36. Inequalities: Theory of Majorization and Its Applications (2011) — link
  37. HyperMultiAssetedADL : A taxonomy of the 10/10 ADL on Hyperliquid (2025) — link
  38. HyperReplay (2025) — link

3 of 46 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2601 linked papers.