Autodeleveraging: Impossibilities and Optimization

Battle Count: 50

Highly relevant for quantitative trading in crypto derivatives markets. The paper directly addresses tail-risk management, liquidation cascades, and loss socialization mechanisms that affect leveraged trading strategies. The ADL trilemma and moral hazard results inform position sizing, leverage decisions, and venue selection for quant traders. The empirical analysis of the October 10, 2025 event ($2.1B liquidations in 12 minutes) provides concrete risk parameters. The fairness and robustness results (RAP, pro-rata) are directly applicable to exchange selection and strategy design. The Stackelberg game framework models the strategic interaction between exchanges and traders during stress events.

The theoretical framework is mathematically sophisticated, involving extreme value theory, majorization theory, convex optimization, game theory, and online learning. The water-filling algorithms (Algorithm 1 and 2) are implementable in linear time. The MDIC controller requires mirror descent with primal-dual updates. The empirical reconstruction pipeline involves two-pass replay, wave clustering, and multiple benchmark allocations (including ILP solvers). Production implementation requires careful handling of contract-space vs wealth-space mappings, integer rounding, and real-time computation under stress conditions. The trilemma proofs and asymptotic results are primarily theoretical.

The paper provides an OSS reproduction pipeline (OSS/out/ directory structure), references a public GitHub repository (ConejoCapital/HyperMultiAssetedADL) for Hyperliquid data reconstruction, and uses publicly available Hyperliquid fill data. However, the theoretical proofs are extensive and the empirical reconstruction relies on partial observability assumptions. The methodology is well-documented with explicit formulas and algorithms (Algorithm 1 and 2 for water-filling).