Full historical L2 depth, every trade and the reference prices, tick by tick, as parquet or CSV or over a REST API. Three lines of pandas and it is a DataFrame. One download covers one market — a series or an event — for one UTC day. Start with 5 free market·days; no card.
One click claims a July 2026 day of a recognisable market — L1, L2, trades and 1-minute candles. Any other market·day is one search away in the catalog; Hyperliquid coins live in their own.
This runs on the free sample as shipped — a full UTC day of the BTC Up or Down 4h series, no account. The numbers on the right are what a market buy would actually have paid on its busiest market, walking the 25-level ladder snapshot by snapshot, against the mid-fill a price-series backtest assumes.
import pandas as pd# the free sample: tickfoundry.com/samples — no accountbook = pd.read_parquet("l2.parquet") # 25 levels a side, every updatetrades = pd.read_parquet("trades.parquet") # every fill the venue reportedtokens = pd.read_csv("reference/tokens.csv", dtype={"token_id": str})book = book.merge(tokens[["token_id", "outcome_label"]], on="token_id")book["ts"] = pd.to_datetime(book.ts_recv_ns, unit="ns", utc=True)book.set_index("ts")[["best_bid", "best_ask", "ask_sz_1"]].resample("1min").last()
A mid-fill model prices a $2,000 order about 5¢ too well, half the time, on a contract near 50¢. Around news it is worse: in the Spain–Belgium quarterfinal the spread went to 31¢ while the mid moved 64 points in 0.65 seconds.
Or skip the code. The TickFoundry Lab replays this exact day tick by tick in your browser (chart, depth ladder, trade tape) and simulates what your order would have paid. Included with every account, no install. Open it in the Lab →
Your edge compounds in strategies, models and papers — none of it compounds in a websocket babysitter. Running your own capture is a standing job that returns nothing to the thing that actually makes money. We already run it; the hours stay yours.
Keep live websockets up around the clock — reconnect, resequence, alert on drops, forever.
Parse the raw feed, de-duplicate, cross-fill gaps and chase schema drift before a single backtest runs.
Stand up storage, retention, monitoring and a 3am page for a system that isn't your product.
A live feed only records what you were watching — last month's market you ignored is gone for good.
Polymarket's feed is real-time only — once a book update scrolls past, it's gone, and no API call brings it back. What a backtest needs from history is fidelity — real fills, real depth, real spread at the instant of the trade, in arrival order. Here's the line-by-line difference.
Start before 15 Oct 2026 and your discount never expires — not when prices rise, not when you move up a tier. Plans started before then also keep full archive access for as long as they run; later plans get a history window from their start date. Applies to both roaming plans. No tier-bumps, no weasel clauses.
Claim founder pricing →Top of book and the 25-level book on every change, and every fill, for every perp, spot pair and HIP-3 market — rebuilt from the chain's own order-event stream, scored against the venue's snapshots and continued every morning. 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 plan roams one venue; the other is a $50/mo add-on.
TickFoundry is Polymarket historical data as a product: tick-by-tick price history, full 25-level L2 order-book depth and the complete trade tape, captured live from the venue since 2026-05-11 and delivered as parquet/CSV downloads, a REST API and SFTP. On-chain trade fills reach further back — back to 2023-08, per market — thinner before 2026 — read from Polygon settlement and sold as their own layer. Since September 2026 the same plant also sells Hyperliquid historical data: top of book, the 25-level order book and every fill for every coin since 2025-01-25, rebuilt from the chain's own order-event stream.
Start with the free tick data sample — a real slice of Polymarket order books and trades in the exact schema paid bundles ship in — then browse the catalog of every market·day on disk, or check the research notes computed from the same files we sell.
Claim 5 free market·days, buy a single market·day from $2 with no subscription, or pick a roaming plan for the whole venue and pull your first archive today, dashboard or API.