{"summary":"Public empirical research on on-chain perpetual-futures microstructure, with Hyperliquid as the substrate. The method is the point: pre-registered experiments, estimators validated against known answers, narrow claims, withdrawn claims kept in the open.","repository":"https://github.com/QuasareumGroup/hyperliquid-microstructure","license":{"code":"MIT","paperAndData":"CC BY 4.0"},"method":["Pre-registered from experiment 17 onward: hypothesis and falsification criterion committed before the run.","Estimators validated against known answers before use.","Withdrawn claims stay public, with the reasons.","Adversarial review by two independent model reviewers; their reports and the response are public."],"publications":[{"slug":"paper-liquidations","title":"Counting fills misrepresents liquidations: a reconstruction unit that has not been priced","author":"Thomas Erhel","status":"preprint","venue":"SSRN","ssrnId":"7189219","doi":"10.2139/ssrn.7189219","version":"2","revised":"2026-08-11","url":"https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7189219","license":"CC BY 4.0","findings":["In Hyperliquid's public fill archive (four fixed UTC hours a day, 27 Jul 2025 – 26 Jul 2026), 351,648 liquidation episodes were reconstructed from 1,005,157 fills.","Counting fills instead of episodes inflates the event count 2.86× (95% CI 2.72–3.00); 3.17× excluding BTC.","The distortion sits in the largest events: below the 90th percentile the median liquidation is a single fill (the mean is 1.49, and that part still carries 46.8% of all fills); in the top 1% the median is 36 fills, and that percentile carries 67.3% of liquidated notional."]},{"slug":"paper-lag","title":"An on-chain venue trails the centralised market by half a second, in every hour measured","author":"Thomas Erhel","status":"working paper (draft, public in the repository)","findings":["On trade prints, Binance leads Hyperliquid by a median 575 ms on BTC, in every one of 191 hours measured (18–26 Jul 2026).","Across 758 asset-hours (BTC, ETH, SOL, HYPE) every hour is positive; the same estimator puts OKX within 25 ms of Binance.","Open: how much is mechanical (block cadence, network) versus price discovery. A timing measurement, not a trading claim."]}],"strategyResearch":"Strategy research stays private."}