Datasets:

aliplayer1
/

polymarket-culture-data

license: mit
task_categories:
  - time-series-forecasting
  - tabular-classification
tags:
  - polymarket
  - prediction-markets
  - culture
  - elon-musk
  - twitter
  - social-media
  - finance
language:
  - en
pretty_name: Polymarket Culture Markets
size_categories:
  - 1M<n<10M
configs:
  - config_name: markets
    data_files:
      - split: train
        path: data/markets.parquet
  - config_name: prices
    data_files:
      - split: train
        path: data/prices/**/*.parquet
  - config_name: ticks
    data_files:
      - split: train
        path: data/ticks/**/*.parquet

Polymarket Culture Markets

Multi-outcome prediction market data from Polymarket for cultural events. Currently covers Elon Musk tweet count markets across 4-day, 7-day, and 1-month timeframes.

Updated automatically every 3 hours.

Subsets

from datasets import load_dataset

markets = load_dataset("aliplayer1/polymarket-culture-data", "markets")
prices = load_dataset("aliplayer1/polymarket-culture-data", "prices")
ticks = load_dataset("aliplayer1/polymarket-culture-data", "ticks")

Or query with DuckDB:

import duckdb

duckdb.sql("""
    SELECT * FROM 'hf://datasets/aliplayer1/polymarket-culture-data/data/prices/**/*.parquet'
    WHERE crypto = 'ELON-TWEETS'
    LIMIT 100
""").show()

Data Description

`markets` — Market metadata

One row per market with question text, resolution, timeframe, and a JSON tokens map of token IDs to outcome labels.

Column	Type	Description
`market_id`	string	Polymarket market identifier
`question`	string	Market question text
`crypto`	string	Event category (e.g. "ELON-TWEETS")
`timeframe`	string	Market timeframe (4-day, 7-day, 1-month)
`volume`	float32	Market volume in USDC
`resolution`	int8	Market resolution
`start_ts`	int64	Market start timestamp (epoch seconds)
`end_ts`	int64	Market end timestamp (epoch seconds)
`condition_id`	string	On-chain condition identifier
`tokens`	string	JSON map of token ID to outcome label

`prices` — Long-format price history

Price time series in long format (one row per token per timestamp), unlike the binary crypto dataset which uses wide format (up_price/down_price columns). Hive-partitioned by crypto and timeframe.

Column	Type	Description
`market_id`	string	Polymarket market identifier
`timestamp`	int64	Price timestamp (epoch seconds)
`token_id`	string	CLOB token identifier
`outcome`	string	Outcome label (e.g. "10-19", "20-29", "Yes")
`price`	float32	Outcome price (0.0-1.0)
`crypto`	string	Event category
`timeframe`	string	Market timeframe

`ticks` — Trade-level fills

Individual trades from on-chain events and WebSocket captures. Same schema as the crypto dataset's ticks table, Hive-partitioned by crypto and timeframe.

Column	Type	Description
`market_id`	string	Polymarket market identifier
`timestamp_ms`	int64	Trade timestamp (epoch milliseconds)
`token_id`	string	CLOB token identifier
`outcome`	string	Outcome label
`side`	string	"BUY" or "SELL" (taker perspective)
`price`	float32	Trade price (0.0-1.0)
`size_usdc`	float32	Trade size in USDC
`tx_hash`	string	Transaction hash ("" for WS ticks)
`block_number`	int32	Polygon block number (0 for WS ticks)
`log_index`	int32	Log index within block
`source`	string	"onchain" or "websocket"
`spot_price_usdt`	float32	Not applicable for culture markets
`spot_price_ts_ms`	int64	Not applicable for culture markets

Pipeline

This dataset is produced by polymarket-data-pipeline. See the crypto dataset (polymarket-crypto-updown) for full pipeline documentation.

License

MIT