a marketmaker.cc project

Prediction Markets

Every trade, wallet and order book across prediction markets, cleaned for research. Wallet PnL that reconciles, wallets sorted by how they trade, bars ready for a backtest, the same outcome priced on every venue. Over REST, WebSocket and Parquet.

Follow a market ↓ scroll to play the explainer

A prediction market ends at one dollar or zero. Everything before that is people paying for a view, and some of them are paid for being right, fast or simply there. We record every fill, know who sent it, and keep the history in a form you can test against.

01 / one market

First fill.
Last dollar.

Scroll. One market from listing to resolution, and what we record along the way.

  • Fills buying YES
  • Fills buying NO
  • Market makers quoting both sides
  • Bars, wallet classes and the result

02 / pipeline

From a block
to a backtest.

01

Collect

Fills from Polymarket, Opinion, Predict.fun, Limitless and Probable, with block, transaction and log index. Order books, markets and events, matched across venues.

02

Reconcile

Splits, merges, neg-risk conversions and payouts are booked, so a wallet's PnL matches what the venue shows. Exchange contracts and service addresses are flagged and left out.

03

Classify

Maker or taker on every fill. Wallets sorted by how they trade: market makers, snipers, longshot buyers, directional traders. A fingerprint finds the same bot when it moves to a new address.

04

Bar

Every token, every minute, hour and day: 75 fields from price, flow and the book to holders, minting, other venues and event context. Values as they were known at the close.

05

Deliver

REST with cursors, a WebSocket by market or wallet list, and daily Parquet dumps. History from 2024, closed and voided markets included.

03 / datasets

Six datasets.
One schema.

Same ids, same units and the same clock on every venue. Join them without a cleanup step.

Trades

Every fill with maker and taker, counterparty, fee, outcome, block and log index. Cursor pages and daily dumps, not offset loops.

trades

Wallets and PnL

Realized and open PnL for any period, by market and by whole event. Service addresses flagged, numbers that match the venue.

wallets

Wallet classes

Market maker, sniper, longshot buyer or directional. Maker share, two-sided markets, holding time and the funding graph behind each wallet.

wallet_class

Order books

Book history with depth at 1, 2 and 5 cents, adds and cancels. Enough to price a copy trade at its real size.

books

Cross-venue

The same outcome on every venue, matched. YES plus NO across venues, the arbitrage gap and its depth, in history.

cross_venue

Crowd vs price

Where the money, the trades and the wallets stand, at fixed times before resolution, against the price at that moment.

majority

04 / bars

One candle has six fields.
Our bar has 75.

Pick the groups you need with fields=. Every bar knows whether anything traded in it, so a flat line is never mistaken for a quiet market.

bar schema · 11 groups75 fields
fills ≥ $500 · all venues

05 / crowd vs price

The crowd is right.
Is it right enough?

A favourite at 0.80 that wins 80% of the time has no edge. The gap between how often a majority wins and what the price said is the signal. Measure the majority by money, by trades or by wallets. Sample data.

hover a point

06 / access

Every fill.
One socket.

Stream trades by market, by event or by a list of wallets. Pull bars, books and wallet history over REST with cursor pagination, or take daily Parquet dumps for the full history.

Prices are in shares from 0 to 1 on every venue. Every record carries the chain or matching time and our receive time in µs.

API preview
// fills from the wallets you follow, on every venue
  const ws = new WebSocket("wss://prediction-markets.marketmaker.cc/api/v1/stream");
  ws.onopen = () => ws.send(JSON.stringify({
    token: "pm_live_•••",
    channels: ["trade", "signal"],
    filter: { wallets: ["0x204f…5e14"], min_usd: 500 },
    resume: "cur_9a2e"   // replay what you missed
  }));
  ws.onmessage = e => console.log(JSON.parse(e.data));
wss://…/api/v1/stream · sample events

07 / rules

Clean enough
to bet on.

No look-ahead

A bar holds what was known at its close and is never rewritten. Results live in their own table, not inside the bar.

One unit

Prices are shares from 0 to 1 on every venue and every endpoint. Sizes in shares and in USD.

Resolution quality

Winning outcome, resolution time, disputes, voids and 50/50 settlements are flagged, so they do not skew your statistics.

Microsecond clock

Block time, matching time where the venue gives it, and our receive time, all in µs.

Test the edge before you pay for it.

Trades, wallets, books and bars across prediction markets, delivered as data.