Replication Materials for a Study of Collateral Architecture, Leverage Choice, and Liquidation Paths in Perpetual Futures

This record contains the frozen replication materials for a study of collateral architecture, leverage choice, liquidation risk, and post-liquidation continuation in perpetual-futures markets. The empirical setting is the October 10–11, 2025 Hyperliquid liquidation cascade. The study examines whether the risk information carried by the same nominal leverage category differs across cross-margin and isolated-margin collateral architectures. It links a complete pre-shock account and position snapshot to reconstructed wallet-coin execution paths, including voluntary position reductions, forced liquidations, first re-entry, and ordered subsequent liquidation. The frozen sample includes: 223,034 raw pre-shock positions; 223,029 positions satisfying the common leverage-validity rule; 54,551 wallets in the common-validity master sample; 54,332 wallets with an eligible reconstructed position at the shock; and 11,474 rebuilt liquidated wallets in the master sample. The deposit contains: the frozen unified analysis specification; final sample-flow, model, and multiple-testing registries; code and aggregate results for wallet and position construction; deterministic liquidation and event-path reconstruction; wallet-level and exact wallet-coin position models; minute-60 competing first-outcome and late-survivor analyses; post-liquidation first-re-entry and ordered subsequent-liquidation analyses; robustness analyses, auto-deleveraging sensitivity checks, and external-event feasibility screens; validation reports, execution logs, input checksums, and SHA-256 file checksums; source materials for the reported tables and figures; a public data dictionary, file manifest, frozen-artifact crosswalk, and raw-data access instructions. Version 2.1.0 updates the record title, documentation, citation information, terminology, and package organization for a derivatives and risk-management presentation of the research. It does not add, remove, or re-estimate any empirical analysis. The frozen sample definitions, statistical models, aggregate result files, and reported numerical results are unchanged from Version 2.0.0. Files inherited from Version 2.0.0 retain their original CURRENT_JOEP_ filenames inside the frozen_analysis_artifacts_v2_0 directory. These legacy filenames are preserved to maintain the original checksums, computational references, and reproducibility of the frozen analysis pipeline; they do not indicate the current journal or manuscript framing. Raw wallet-level archives and wallet addresses are not redistributed in this deposit. The underlying Hyperliquid archives are publicly accessible through requester-pays object storage and official node-fill archives. Access requirements and reconstruction procedures are documented in CURRENT_JDQS_raw_data_access_instructions_2026-08-03.md. The analyses estimate conditional associations. Margin architecture and leverage choice are self-selected, and the materials do not identify causal effects or traders’ underlying motives. A wallet may represent an individual, an institution, or an automated trading strategy. For package contents, file-level descriptions, integrity checks, and the recommended reproduction sequence, see: README.md CURRENT_JDQS_file_manifest_2026-08-03.csv CURRENT_JDQS_frozen_artifact_crosswalk_2026-08-03.csv CURRENT_JDQS_raw_data_access_instructions_2026-08-03.md CURRENT_JDQS_SHA256SUMS_2026-08-03.txt CURRENT_JDQS_version_2_1_integrity_report_2026-08-03.md