Dataset

One year of Polymarket trades, already joined to outcomes

Raw trades are public — anyone can pull them from the API. What is hard to obtain is a continuous year of them, matched to how each market actually resolved. That join is the work, and it cannot be done retroactively: Polymarket's pagination limits stop you from rebuilding a year of history after the fact. We have been collecting continuously since July 2025.

14,044,654

trades

42,683

resolved markets

12 months

Jul 2025 → Jul 2026

Free sample — 10,000 rows

Evenly sampled across the whole year, not the first N rows. Same schema as the full set. No signup, no email required.

Download sample (CSV, 2.1 MB)

Schema

Column	Meaning
ts / date_utc	Trade timestamp (unix + ISO)
conditionId	Market identifier, joinable to Polymarket
question	Market question as published
category	Normalised category (crypto, politics, soccer, …)
side	BUY or SELL
outcome / outcomeIndex	Which contract was traded
price	Execution price, 0–1
size / usd	Contracts and notional value
winnerIndex	Which outcome actually won
won	1 if this trade backed the winner, 0 otherwise
nOutcomes	Number of outcomes in the market
marketEndDate	Resolution date

What it is good for

Testing a strategy against outcomes that are already known, on a sample large enough to split out-of-sample.
Measuring the favourite–longshot bias, category effects, or price-band behaviour on real money rather than simulation.
Research on prediction-market efficiency with a full year of one venue.
Building your own tool without spending a year collecting first.

Full access

The complete set is 14M rows. Available as a one-off CSV/Parquet delivery, or as ongoing access that includes newly resolved markets. Tell us what you need it for and we will quote — academic and non-commercial use is priced differently.

Note on provenance: every row derives from Polymarket's public on-chain trade record. We add collection, normalisation and the outcome join — we make no claim of exclusivity over the underlying facts.

[email protected]

Curious what we found in it? We published the analysis — including the strategies that did not work.