Quantitative research on on-chain market microstructure — Paris, France.
We study perpetual-futures markets on their complete public records, with the discipline the results need to be believed: experiments pre-registered before they run, retractions kept in the open, estimators validated against known answers before use.
Current programme — hyperliquid-microstructure
24 pre-registered experiments on Hyperliquid's public data. Established so far:
| result | number |
|---|---|
| Counting liquidations from fill feeds inflates event counts | 5.72× (95% CI [5.47, 5.99]) — and the error concentrates on the largest events |
| Binance leads Hyperliquid price discovery | ~575 ms, in 100% of 758 measured asset-hours, zero reversals |
The first paper — Counting fills misrepresents liquidations: evidence from a complete on-chain derivatives ledger — is under review at SSRN. The PDF, every dataset, the analysis code, and two independent adversarial model reviews with the point-by-point response are already public in the repository.
More than a dozen claims have been withdrawn or requalified along the way — each correction stays in the file that made the original claim, with the reason. The repository's method rules were each paid for by a real error, and the git history shows the failures as well as the results. That is deliberate: work that only shows successes cannot be audited.
perplog — multi-venue orderflow capture and analytics for Hyperliquid. The research above runs on its tape; the tape is validated by the research.
Contact: contact@quasareum.com
