Replication package: Liquidation-Volume Measurement Error — Calibrating Two Substitutes against Hyperliquid's Native Tagged Record
Year: 2026
Type: software
Abstract: Replication package for the JFDS submission "Liquidation-Volume Measurement Error: Calibrating Two Substitutes against Hyperliquid's Native Tagged Record". Supersedes v9. Change in this version: the latest-kept throttle replay now keeps the last execution matched in a block. Hyperliquid stamps fills with block times, so 93-100% of occupied one-second buckets contain a single timestamp and the rule is decided by the order of execution within the block; the raw archive lists a block's executions in matching order (the liquidated account's running position chains forward in 191,372 of 191,372 consecutive pairs), and v9 had kept the first execution of the block, the earliest rather than the latest. exp1_exp3d_scan.py gains a tie-break parameter and re-runs every latest-rule variant and the bucket-origin sensitivity under the last-arrival reading (new keys latest_tie_last and *_tielast; all v9 keys unchanged). A new tie audit (exp3f_tie_audit.py) re-scores the rule under five orderings of the tied candidates, asserts that its first- and last-arrival figures reproduce exp3d exactly, verifies the within-block order from the liquidated account's running position and from counterparties' positions across different liquidated accounts, measures how far such account chains reach within a block (they rarely reach the ADL executions, whose end-of-block position rests on the archive order itself), records where ADL executions sit inside their blocks, and records the size and channel of the first- and last-arrival picks. make_exhibits.py (Exhibits 6 and 7C), the figures and the no-regression gate are updated accordingly. Four files are added and eight modified relative to v9. All numbers regenerate from the recorded runs; the only random seed is 20251018. The three platform-wide ADL reconciliation figures quoted in the text ($2.103 billion, 35,022 executions, 162 tickers) come from the first author's dissertation extraction and are not reproduced by this package; the text states this. Derives from the dataset archived at DOI 10.5281/zenodo.18759046.
Author: LUO, YAN
Language: English
Subfield: Finance
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W7204228722
