Explore the data before you write a line of code: replay any market·day tick by tick with a synchronized chart, depth ladder and trade tape, step through exact timestamps and quality flags, and simulate what a buy or sell of N shares would have cost against the recorded book. No extra charge.
Coming soon: curated research recipes (spread and depth into expiry, slippage curves by order size, Hyperliquid liquidity), a Python SDK for Lab runs, and more venues in the workbench.
Open the Lab on the public sample day →Every plan is the whole venue and one named seat. The plans differ on book depth, history, API rate and delivery; nothing is priced per market. A plan started at the founder rate keeps full archive access for as long as it runs.
Book + tick capture from 2026-05-11; on-chain trade fills back to 2023-08, per market — thinner before 2026, read from Polygon settlement and requested separately from market·day bundles. Prices in USD before VAT: EU customers see their country's VAT at checkout unless they enter a valid VAT ID (reverse charge).
Every plan above is the whole venue, so the only question left is whether the market you actually care about is in it, and how far back it runs. Search by title, ticker or team; answers come from the same catalog the bundles are built from.
That is the whole plant on cloud pricing — redundant capture, storage that only grows, the compute that turns raw frames into usable data — billed every month, mostly for markets you're not even sure you'll use. We run all of it; you pull what you need and your time goes to backtesting and trading.
Hyperliquid is starker still. Building this yourself — pulling the chain's order-event archive out of requester-pays S3 and replaying it into books — runs to around $10,000 in egress and compute. A month of either plan is a rounding error on that.
There is no historical endpoint. A day you did not capture is gone at any price. Our capture runs gapless from 2026-05-11.
over 350,000 outcome tokens, and the upstream drops more than a handful of connections per IP. The subscribe list has to fan out across a fleet before you get one tick.
A lone collector holes on every reconnect, so it has to run redundantly and be watched. An outage at 03:00 is a day nobody can re-record. No server quote includes the pager.
Storage only grows, and you would have to pick today what to keep. Anything you skip cannot be captured back. Here you pull what you use and leave the rest on our disks.
Capture, monitoring, reconstruction and storage are all handled, and the 3am pager is ours. You get finished parquet, ready to backtest, from 2026-05-11. Premium is $149.25/mo at the founder rate, which is less than the engine alone would cost you to run.
how we verify the dataThe feeds the markets themselves resolve against — not the order book, but the inputs. Polymarket's real-time data stream captured continuously and normalized to one whole file per UTC day, every symbol and game included. Sold per day from the catalog; a free account gets all of 2026-07-01, both feeds, without spending a market·day claim.
Exchange spot, Chainlink oracle, and the 30s/60s TWAPs that settle the crypto up/down markets — the reference series a contract actually resolves on. ~2.5M rows/day.
Score, period, status and lifecycle for every sports market, sampled continuously, with the full upstream state object preserved per row — the ground truth a sports market reprices against.
Coverage since 2026-06-21, gapless. Not all topics span it: the market-resolving TWAPs start 2026-08-04, and the equities/FX/metals topic is a closed era that ended 2026-07-30 when upstream discontinued it — real, sellable data, but no new days. Every bundle's README states exactly which topics your chosen range contains.
Polymarket's perpetual futures venue — 67 instruments across crypto, US equities, indices and commodities: event-driven BBO, 20-level books every ~100 ms, the full ladder every ~15 s, trades, and tickers with funding and open interest. Sold per day with an instrument filter from the perps catalog; a free account pulls one fixed sample day (2026-08-12) at full fidelity.
Captured on two collectors since 2026-07-31 (a partial first day). Books are full snapshots rather than deltas, funding and open interest live in the ticker table, and every bundle's README states the caveats for your exact range — see the perps schema.
The cross-venue perpetuals package — Polymarket's perps exchange (books, BBO, trades, funding) and every Hyperliquid perpetual, native and HIP-3, on the venues' own clocks. Same-symbol pairs for crypto natively and for equities, indices and commodities on the xyz: dex. Hyperliquid spot pairs are not included: for those, buy the plan for Hyperliquid. Polymarket's prediction markets can be added for $50/mo. Pull from the perps catalog and the Hyperliquid catalog.
Why one package: the two venues list the same names (BTC, ETH, SOL natively; NVDA, TSLA, GOLD, the S&P 500 on Hyperliquid's xyz dex) on their own clocks, and research note 006 measures Polymarket printing 250–500 ms ahead of Hyperliquid on source timestamps. Start from perps tick data.
Every Hyperliquid perp, spot pair and HIP-3 market since January 2025 — top of book on every change, the 25-level order book on every change, and every trade, reconstructed from the chain's own order-event stream rather than sampled from a websocket. Sold per day with a coin filter from the Hyperliquid catalog; a free account pulls the fixed sample day (2026-08-12, any coins, up to 1.0 GB). The same rule runs the other way: a plan bought for Hyperliquid adds all of Polymarket for $50/mo.
Coverage from 2025-01-25, continued daily; your plan's history window applies here as on Polymarket. Full 25-level books on every plan — Explorer and Premium differ in history and limits, not in the file. Every bundle carries a per-coin quality scorecard and a README with the caveats for your exact range — see the Hyperliquid schema.
Pull a single market for one date — no subscription. For spot-checking data quality or reproducing one event. Files in about a minute.
Browse catalog →