For data buyers

Prediction market data you can build on.

Every Polymarket trade and every wallet, labeled and checked against the chain. Every trader's P&L rebuilt from raw chain data and set beside Polymarket's own figure. Labels stored as they were each day, documented field by field.

APITM
Market makers
Who is quoting both sides, with the score and the sample behind the label. Refreshed daily at 08:00 UTC.
FieldTypeWhat it is
addressstringLowercase 0x wallet.
mm_classstringThe stored label.
mm_scorenumberThe score behind the label, 0 to 100.
mm_versionstringThe model that produced it, like mm_v1.2.
fillsintegerFills the score was measured on.
active_daysintegerActive days in that sample.
classification_changed_atstringWhen any label on the row last changed.
Nov 2022Every Polymarket trade since, decoded from raw chain data
16Classifiers on every wallet
MatchedEvery trader's P&L checked against Polymarket's own figure
What a data team gets

Complete, documented,
and checked against the chain.

When we went looking for Polymarket data, what existed was often wrong. The P&L didn't match the chain, and the data stopped at the surface. So we decoded it ourselves. You get the history you need to defend a model, labels that say who's behind every trade, and a schema you can read before you sign.

Coverage

Decoded from the chain, not scraped.

Every Polymarket trade since November 2022, read straight from raw chain data. That includes the splits, merges, redemptions, and transfers the venue's own feed never shows, which is how every trader's P&L can be rebuilt and checked.

  • SourcePublic chain data, decoded by TAPE
  • HistorySince November 2022
Coverage
Polymarket, as of November 2026
Trades2.18B
Wallets3.3M
Markets5.87M
SinceNov 2022
What's decoded From raw chain data
TradesIncludedEvery fill on the venue
Splits and mergesIncludedNot in the venue's feed
RedemptionsIncludedHow winnings are paid out
TransfersIncludedPositions moved between wallets
Delivery

Delivered where your stack lives.

Use the terminal, call the API, or query the data in your own Snowflake through the datashare. Your AI tools can reach it through the MCP server.

  • WorkstationThe terminal
  • APIJSON over HTTPS
  • Snowflake datashareIn your warehouse
  • MCPFor your AI tools
Four ways in
The same data behind each
WorkstationThe terminal, for analysts and traders.
APIJSON over HTTPS, documented field by field.GET /v1/market-makers
DatashareThe warehouse, delivered into your Snowflake.select * from …
MCPAsk from your own AI assistant.
Derived data

History a backtest can trust.

Sixteen classifiers on every wallet, trader skill in 31 categories and by bet type, and P&L reconciled against Polymarket. Every label is stored as it was each day, with the model version that produced it. So you can test a strategy on what was known at the time, not on what's known now, and a change shows up as a new dated row while the history stays as it was.

  • Classifiers16
  • StoredDaily, since June 29, 2026
  • VersionedModel version on every label
Label history
0x7d…a3f0 · market making, as stored each day
DateLabelScoreModel
2026-11-03Market maker81.4mm_v1.2
2026-11-02Market maker80.9mm_v1.2
2026-11-01Mixed63.0mm_v1.2
2026-10-31Mixed61.7mm_v1.2
2026-10-30Mixed60.2mm_v1.2
Changed on 2026-11-02 The earlier rows are kept as they were
Freshness

Every dataset says how fresh it is.

Every dataset reports its newest row and how often it refreshes, from hourly to daily. When a read fails, it says so, instead of handing back an empty answer that looks like no data.

  • Shown/v1/meta/freshness
  • CadenceHourly to daily, by dataset
GET /v1/meta/freshness
Illustrative response
{
  "datasets": [
    { "name": "Market makers", "as_of_status": "measured",
      "refresh": "Daily, with the classifier build (08:00 UTC)" },
    { "name": "Copy leaders", "as_of_status": "measured",
      "refresh": "Daily at 10:00 UTC" },
    { "name": "Events", "as_of_status": "measured",
      "refresh": "Hourly" },
    { "name": "Trades", "as_of_status": "measured",
      "refresh": "Hourly at :20" },
    …
  ]
}
What's labeled

Every wallet, labeled seven ways.

The same trade reads differently once you know who placed it. On the chain, a wallet bought 12,400 Yes shares at 62¢. In TAPE, a top 5% NBA trader on a hot streak added to a winning position. These are the label families behind that read.

FamilyWhat it tells you
Trader typeMarket maker, directional, high frequency, or retail.
Trading styleHow long they hold, from scalper to position, how they exit, how they size, how widely they trade, and when.
LifecycleHow long they've traded, how active they are, whether their form is rising, steady, or cooling, and whether the wallet is fresh.
MoneyWhere the funding came from, money in and out of Polymarket, and what the venue pays them.
PerformanceSkill in 31 categories and by bet type, with P&L checked against the chain.
NetworkWho gets copied, how much money follows, and how fast.
MarketsCategory, bet type, holders, depth, spread, and settlement risk.
History and freshness

What we have, and since when.

Your legal and engineering teams will ask. Here it is up front.

Nov 2022 Jun 29, 2026 Today
Every tradeLabels as they were each day
  • TradesEvery Polymarket trade since November 2022, decoded from raw chain data.
  • LabelsStored as they were each day since June 29, 2026, with the model version on every row.
  • SourcePublic chain data, decoded by TAPE.
  • RefreshEvery dataset reports when it was last updated and how often it refreshes, from hourly to daily.

The catalogue

Every data set that is live today, what one row is, how far back it goes, and how it reaches you.

Data setOne row isHistoryRefreshDelivered by
Trade historyOne fill on Polymarket, decoded from the chain.Since November 2022Continuously, usually within the hour of the chainDatashare
TradesOne trading decision, with every fill it took to complete.Since September 2, 2026. Decisions over $1,000 are kept for good, the rest for 180 days.Hourly, up to about 90 minutes behind the venueAPI /v1/trades
TradersOne wallet: its verdict, skill, risk, flow, and where its P&L came from.Today's recordDaily, after the 08:00 UTC classifier buildAPI /v1/traders
Market makersOne wallet, with its market maker label and the score and sample behind it.Today's labelsDaily, after the 08:00 UTC classifier buildAPI /v1/market-makers
Copy leadersOne wallet that others copy: how many copy it, and what the price does after it trades.Today's snapshotDaily at 10:00 UTCAPI /v1/copy-leaders
EventsOne game, match or series, with its markets resolving in the next 30 days.Upcoming eventsHourlyAPI /v1/events
FreshnessOne data set, with its newest row and how often it refreshes.NowOn every requestAPI /v1/meta/freshness
Before the demo

Questions data teams ask first.

Is it complete?

The trade history is every Polymarket trade since November 2022, decoded from the chain. A few datasets cover a defined group of wallets instead, like the Smart Money Index and the traders who qualify for it. We'll show you which is which.

What's the schema, and does it change?

Every field is documented with its type and what it means. Labels carry the version of the model that produced them, so a change to a model is visible in the data rather than hidden in it.

What are the license and redistribution terms?

We'll take you through them in the demo.

How fresh is it, and what happens when a load fails?

Every dataset reports its newest row and how often it refreshes. If a read fails or runs late, the response says so, so a gap never looks like an empty dataset.

Where does the data come from?

Public chain data, which TAPE decodes. Nothing is scraped.

Can we evaluate it before we sign?

Talk to us. We'll work out what your team needs to see.

See the schema before you sign.

Book a demo and we'll walk your data team through it.