← Back to blog

ZenHodl vs PolymarketData: Polymarket Historical Data

updated 2026-10-08 comparison data polymarket kalshi polymarketdata

By ZenHodl. Dataset documentation, research and model evaluations are linked in the article. A separate filtered ledger of bot-attributed trades, including losses and its admission rules, is public at /results.

Choose the data source around the question you need to answer. Polymarket history queried through an API, sampled Kalshi depth and a fixed matched-MLB panel are different products. Their row totals cannot be compared without checking what one row represents.

We checked public vendor pages and ZenHodl's catalog on October 8, 2026. We did not inspect paid PolymarketData payloads or independently measure its collection completeness.

PolymarketData's documented offering

PolymarketData's homepage advertises historical prices, market metrics and L2 snapshots from August 2025, with one-minute resolution and more than one million markets. It offers a REST API, Python SDK, S3 delivery in Parquet/CSV/JSON and custom category or time-range exports. These are the vendor's descriptions, not our independently verified coverage figures.

The homepage explicitly permits one-time S3 dumps as well as scheduled exports. Our earlier “subscription-only” comparison was incorrect. The pricing page names Trader, Pro and Ultra self-serve plans and custom Enterprise/Bulk delivery. It states that custom pipelines, derived datasets and extra support are separately priced. The numeric plan cards were not available in our source capture, so we removed older dollar amounts and free-plan limits rather than presenting them as current.

Ask the provider for the exact market list, channel clocks, missing intervals, export license and quote-size fields needed for your analysis. Its public pages do not establish a verified Kalshi dataset in this comparison.

ZenHodl's dated coverage

The public catalog, vintage October 5, lists separate archives with product-specific counts, observation bounds and limitations. Examples:

Product Release figures What the figures support
Polymarket Sports Archive October 3 release: 1,041,590,302 quote rows plus 587,197,084 depth rows Analysis of sampled sports book states; not all exchange updates, all sports events or executed trades
Kalshi Microstructure Tape October 2 release: 821,979,759 depth rows plus 29,984,492 captured deduplicated prints Selected-market depth and print research with the disclosed polling and retrieval limits
MLB Cross-Venue Matched Book 1,394,808 rows; 1,005 dated games, of which 965 are settled and 40 unresolved A derived quote panel with an explicit matching tolerance; not a fill or arbitrage ledger

The older $150 Polymarket/Kalshi archive is a different December–January release with different venue windows. It should not be described as the recent sports archive or the MLB matched panel.

Our Polymarket sports collection runs on roughly 30-second cycles. Cached websocket state can be up to 300 seconds old without an exact-age column. Kalshi depth records a volume-selected subset, up to 20 levels per side. Neither is a complete exchange-event replay.

Game context is product-specific

Some legacy Kalshi rows carry score, period and clock recorded at capture or retrieval time. Legacy Polymarket rows do not share those fields. The MLB matched product derives a separate game-linked panel. Its nearest-time quote join can use a later quote within ±30 seconds; an execution backtest must reject future joins or build a causal match.

Use the linked schemas to establish which fields exist in the product you are buying. “Score synced” without a clock definition is insufficient for event-time research.

Compare delivery and licenses fairly

ZenHodl archives cost $9–$199 for a fixed downloadable release under our data license. An archive purchase does not include every subsequent release or unrestricted redistribution. The live API is a separate product, and paid API access does not include the archive ZIPs.

Because PolymarketData also documents one-time exports, compare quotes for the same dates, venue, channels and permitted use. A file versus API distinction alone no longer separates the services.

Start with ZenHodl samples, then compare the same fields and dates in a provider sample. Record missing data, timestamp semantics and settlement coverage before computing strategy returns.

Sources: PolymarketData homepage, pricing, and ZenHodl catalog, checked October 8, 2026. Historical counts remain tied to the named release rather than a current completeness guarantee.

Related reading

Get ZenHodl Weekly

Dataset releases, research notes, and public results, including corrections.

Research and dataset updates from ZenHodl.

Want the data behind this post?

Historical sports prediction-market datasets with measured coverage, documented schemas, and disclosed gaps.