The same captured tape — every order, every book update, every fill — feeds a wide span of work, from a solo backtest to a desk's reproduction audit. Each job below shows the data layer it needs and the plan that unlocks it.
Every print and every top-of-book move, in arrival order, with nanosecond receive timestamps. Replay a strategy against the real tape — no resampled candles, no survivorship gaps — and measure edge against what was actually quotable at the touch.
Walk the full 25-level L2 ladder to size a fill the way the book really clears. Reconstruct VWAP against resting depth, study queue dynamics and book pressure, and cost a strategy honestly instead of assuming touch-price fills.
Onchain trade fills carry the settling transaction_hash, so every print joins back to a wallet. Follow large or repeat actors across markets, cluster cohorts, and study how informed flow leads price — with onchain trade history back to July 2025, backfilling toward each market's inception (~3 years).
The Binance and Chainlink spot (RTDS) feeds ship aligned with Polymarket on the same receive clock. Quantify how spot moves lead — or lag — the implied probability on crypto markets, and build cross-asset signals on a single timeline.
Full-depth snapshots expose resting size at every level, spread regimes and how the book refills after it's swept. Model adverse selection, inventory risk and quoting behaviour across the whole venue rather than a single market.
Each row carries the collector's nanosecond receive timestamp, so you can measure reaction latency to the tick. Anchor on a resolving event or a headline and watch the book and tape respond in real, replayable time.
Pull at scale over the REST API or SFTP — clean, columnar parquet with stable schemas and join keys, ready to stream into a training pipeline. No scraping, no normalization, no per-market plumbing to maintain.
Raw websocket atoms plus the deterministic replay engine let you rebuild any book at any nanosecond, in-house, byte-for-byte. The provenance trail funds and auditors need when a number has to be defensible.
We're writing up reproducible studies — microstructure, onchain flow and spot-versus-prediction signal — with the notebooks and the exact bundles behind them, so you can rerun every result yourself. They're not published yet; want an early look or to compare against your own work?