Polymarket Historical Data

10M+ price snapshots. 9,550+ markets. 30 days rolling. 15-minute resolution.

The complete dataset behind a public live-traded strategy (302 trades, 79.8% win rate). Free samples on Hugging Face. Full dataset and live API on Gumroad / api.protodex.io.

Coverage

Table	Rows	Description
markets	9,550+	Question, category, volume_24h, liquidity, end_date
prices	10,000,000+	15-min OHLC snapshots for YES/NO outcomes
orderbooks	800,000+	Bid/ask depth snapshots

• Source: Polymarket Gamma + CLOB APIs (no scraping, no proprietary data)
• Update cadence: every 15 minutes via ForgeOS launchd job
• Categories: politics, sports, crypto, economics, geopolitics, weather, science
• Format: SQLite (single file, queryable from any language)

Get the data

Tier	Price	Format	Where
Free sample (1 day)	$0	SQLite + CSV	Hugging Face
Cross-signal sample (BTC/ETH/SOL + Polymarket probabilities)	$0	CSV	Hugging Face
Sample paid (1 day full SQLite)	$1	SQLite	Gumroad
Full dataset (30 days)	$9	SQLite	Gumroad
Live API (no download, query directly)	Free 100/day · $19/mo Pro	HTTP/JSON	api.protodex.io
Live subscription (auto-refreshing dataset)	$29/mo	SQLite	Gumroad

Quick start (Python)

After downloading the SQLite file:

import sqlite3, pandas as pd

con = sqlite3.connect("polymarket.db")

# Top markets by 24h volume
df = pd.read_sql("""
  SELECT id, question, category, volume_24h, liquidity, end_date
  FROM markets
  WHERE active = 1
  ORDER BY volume_24h DESC
  LIMIT 20
""", con)

# 15-min prices for one market
prices = pd.read_sql("""
  SELECT outcome, price, ts
  FROM prices
  WHERE market_id = ?
  ORDER BY ts
""", con, params=("0x...",))

Schema

CREATE TABLE markets (
    id TEXT PRIMARY KEY,
    slug TEXT,
    question TEXT,
    category TEXT,
    volume_24h REAL,
    liquidity REAL,
    end_date TEXT,
    active INTEGER
);

CREATE TABLE prices (
    market_id TEXT,
    outcome TEXT,         -- 'YES' or 'NO'
    price REAL,           -- 0.0 to 1.0
    ts TIMESTAMP,
    PRIMARY KEY (market_id, outcome, ts)
);

CREATE TABLE orderbooks (
    market_id TEXT,
    bids TEXT,            -- JSON array
    asks TEXT,            -- JSON array
    spread REAL,
    ts TIMESTAMP
);

Strategy / research using this data

This dataset powers a public-audited crash-recovery bot:

• 302 live trades · 79.8% win rate · 2.6× win/loss ratio
• Backtest: 6,225 trades on the full dataset, 75% WR
• Documented finding: after a >20% crash, average bounce is +6.6% within 15 minutes (n=5,629)

Full methodology: polymarket-crash-bot. Live PnL: api.protodex.io/live.

What you can build with this

• Backtest any prediction-market strategy without paying $24K/yr for Bloomberg
• Train ML models on prediction-market mispricing signals
• Build dashboards and screeners (see polyscope)
• Feed an LLM agent with prediction-market context (MCP server coming)

License

MIT. Use it, fork it, ship something.

Contact

LuciferForge@proton.me · @gmanjuu · api.protodex.io