antontuzovAI/polymarket-clean-trades-linked-1m

Polymarket Clean Trades Linked 1M

This dataset is a cleaned derivative of:

SII-WANGZJ/Polymarket_data

Main source file:

trades.parquet

Metadata source file:

markets.parquet

Build details

• Rows read from source trades: 1,025,000
• Rows written after cleaning: 1,000,000
• Target clean rows: 1000000
• Max raw rows: 1500000
• Parquet shards: 9
• Approximate local size: 77.77 MB
• Drop unmatched markets: True
• Require valid side: True
• Min price: 0.0
• Max price: 1.0
• Min amount: 0.0
• Remove duplicates: True
• Global dedup limit: 1200000
• Seen trade uid count: 1,024,759
• Remove junk rows: True

Cleaning rules

This dataset filters:

• missing market key
• missing timestamp
• missing price
• missing amount
• prices outside [MIN_PRICE, MAX_PRICE]
• amounts not greater than MIN_AMOUNT
• timestamps outside years MIN_TIMESTAMP_YEAR to MAX_TIMESTAMP_FUTURE_YEAR
• duplicate trades when enabled
• obvious junk/test/demo/admin/spam/placeholder rows when enabled

Column mapping used

This tailored build maps the observed trade columns:

token_amount -> amount
usd_amount -> notional
taker_direction / maker_direction -> side
nonusdc_side -> outcome
asset_id -> token_id / asset_id
transaction_hash + log_index + asset_id -> trade_uid
condition_id -> market_key

Market metadata is linked using:

trades.condition_id -> markets.condition_id / market_key

Schema

Column	Type	Description
trade_uid	string	Composite trade id
market_key	string	Linked market key
market_id	string	Original market id
condition_id	string	Original condition id
token_id	string	Outcome token id / asset id
asset_id	string	Original asset id
timestamp_ms	int64	Trade timestamp in Unix milliseconds
trade_date	string	UTC date
trade_hour	int32	UTC hour
price	float64	Outcome price / probability
amount	float64	Token amount / shares
notional	float64	USD notional
side	string	buy/sell/unknown
maker_direction	string	Original maker direction
taker_direction	string	Original taker direction
maker	string	Maker address
taker	string	Taker address
tx_hash	string	Transaction hash
log_index	int64	Log index
outcome	string	Outcome / nonUSDC side
has_market_metadata	bool	Whether market metadata was joined
market_question	string	Market question/title
market_slug	string	Market slug
market_category	string	Market category/tags
market_end_date	string	Market end date
market_active	string	Active flag
market_resolved	string	Resolved/status
market_outcomes	string	Outcomes serialized as JSON/string
market_outcome_prices	string	Outcome prices serialized as JSON/string
market_volume	float64	Market volume
market_liquidity	float64	Market liquidity
market_description	string	Market description
source_repo	string	Source Hugging Face repo
source_file	string	Source file

Usage

from datasets import load_dataset

ds = load_dataset(
    "antontuzovAI/polymarket-clean-trades-linked-1m",
    streaming=True,
    split="train"
)

for sample in ds:
    print(sample["market_key"])
    print(sample["timestamp_ms"])
    print(sample["price"])
    print(sample["amount"])
    print(sample["notional"])
    print(sample["side"])
    print(sample["market_question"])
    break

Source

This dataset is derived from:

SII-WANGZJ/Polymarket_data

License and usage notice

You must respect the license and terms of the source dataset and any underlying Polymarket/public data sources.

This is a derived cleaned dataset, not an official Polymarket dataset.