Predict.fun

Coverage start date: 2026-09-10.

Available data

Data type	Contents
`l2`	Order book of one market for one UTC day, from the Predict.fun WebSocket. Every row is a complete aggregated book.
`onchain_fills`	Fills of one market for one UTC day, recorded on BNB Smart Chain, with maker and taker wallets.
`lifecycle`	Market status, trading status and market changes for every market, one file per UTC day, from the Predict.fun WebSocket.

There is no trades data type: Predict.fun's WebSocket has no public trade channel; fills come from the chain as onchain_fills. See the Predict.fun L2, on-chain fills and lifecycle schemas for column definitions.

API access and data format

Send your API key in the api_key request header. Get a free key at pmdata.dev/app.

Each file costs one download. A file that does not exist returns 404 and costs nothing.

Predict.fun is a separate product from Plus: Polymarket. Free accounts can download Predict.fun data older than 3 weeks (21 days): 600 downloads at sign-up, then 50 per month, counted separately from Polymarket. Data from the last 3 weeks needs Plus: Predict.fun; Plus: Polymarket does not include it. Plus: Predict.fun also includes Chainlink data.

Market API

Download one market's Parquet file for one data type and UTC date. market_id is Predict.fun's numeric market id. Markets and dates PMData has recorded: see the markets catalog.

GEThttps://api.pmdata.dev/v1/predict_fun/{data_type}/{YYYY}/{MM}/{DD}/{market_id}.parquet

Response schema by data_type: l2 · onchain_fills

import pandas as pd

market_id = 123  # Predict.fun numeric market id
date = "2026/10/01"  # UTC date, YYYY/MM/DD
data_type = "l2"  # l2, onchain_fills

url = (
    f"https://api.pmdata.dev/v1/predict_fun/{data_type}/"
    f"{date}/{market_id}.parquet"
)

df = pd.read_parquet(
    url,
    storage_options={
        "api_key": "<YOUR_API_KEY>",
        "User-Agent": "Mozilla/5.0",
    },
)
print(df.head())

Each L2 file holds one book, priced in the market's first outcome; the other outcome's book is its complement. The outcome parameter is not supported and returns 400.

Market lifecycle API

Download one UTC day of lifecycle updates for every market:

GEThttps://api.pmdata.dev/v1/predict_fun/lifecycle/{YYYY}/{MM}/{DD}/markets_lifecycle.parquet

Response schema: Predict.fun lifecycle

import pandas as pd

date = "2026/10/01"  # UTC date, YYYY/MM/DD

url = f"https://api.pmdata.dev/v1/predict_fun/lifecycle/{date}/markets_lifecycle.parquet"

df = pd.read_parquet(
    url,
    storage_options={
        "api_key": "<YOUR_API_KEY>",
        "User-Agent": "Mozilla/5.0",
    },
)
print(df.head())

Captured real-time channels

PMData collects these topics from Predict.fun's WebSocket, wss://ws.predict.fun/ws:

predictOrderbook/{marketId} — Complete aggregated order book for one market, sent on subscription and on every change. Stored in l2.
predictMarketStatus/{marketId} — Market status, such as REGISTERED or RESOLVED. Stored in lifecycle.
predictTradingStatus/{marketId} — Trading status, such as OPEN or CLOSED. Stored in lifecycle.
predictMarketChanged/{marketId} — Changes to a market, such as the opening and closing prices of crypto Up/Down markets. Stored in lifecycle.

Every market is collected over two independent connections. Each row keeps the source timestamp and PMData's receive timestamp, local_timestamp. On-chain fills are read from BNB Smart Chain, not the WebSocket, and stored in onchain_fills.

Market data collection details

PMData's market data collection infrastructure for Predict.fun is located in AWS ap-northeast-1 (Tokyo), colocated with Predict.fun's exchange.