hftbacktest

This is the machinery a microstructure strategy needs before its backtest means anything. The engine reconstructs a full order book from Level-2 market-by-price and Level-3 market-by-order feeds, replays it tick by tick — on a customizable time interval or driven by feed and order receipt — and simulates fills against a queue-position model, with feed latency and order latency modelled separately and both replaceable with your own. Multi-asset and multi-exchange setups are supported. The Python API is written to run inside Numba JIT functions; a Rust core backs it. The README is explicit about why the detail matters: a simulator should neither be so pessimistic that it hides small edges nor so optimistic that it invents them.

Live deployment reuses the same algorithm code, currently for Binance Futures and Bybit, and is Rust-only by the project’s own feature list — the Python API does not carry it. No market data ships with the package, so preparing tick data is the user’s first task; the documentation covers data preparation, custom data integration and depth fusion, alongside tutorials on grid trading and order-book-imbalance market making. The two distributions version independently: PyPI hftbacktest is at 2.4.4 and the hftbacktest crate at 0.9.4, both published 2025-12-10.

Two caveats about provenance and pace. The maintainer is pseudonymous — the GitHub profile nkaz001 names no person or organization, and the MIT LICENSE gives the copyright holder as an email address — so there is a stable, long-lived identity behind the code but no verifiable one. And the repository has been quiet since 2025-12-23, roughly eight months as of 2026-08-17, after a heavy stretch of commits through the second half of 2025.