valctrl/politics1m

---
license: mit
task_categories:
- time-series-forecasting
- tabular-classification
tags:
- prediction-markets
- finance
- politics
- time-series
- ohlc
size_categories:
- 100K<n<1M
---

# Politics1M Dataset

OHLC price data for political prediction markets, aggregated into 5-minute candles.

## Dataset Structure

- **Format**: Parquet
- **Time Resolution**: 5-minute intervals
- **Source**: Polymarket prediction markets
- **Scope**: Political markets with resolved outcomes

## Data Schema

- `condition_id`: Market condition identifier
- `market_id`: Market identifier
- `token_id`: Outcome token identifier
- `bucket_start_ts`: Unix timestamp (5-minute bucket start)
- `open`, `high`, `low`, `close`: Price values (0-1 range)
- `volume`: Total trading volume
- `trade_count`: Number of trades in bucket
- `question`: Market question text
- `theme`: Market category
- `target_resolution`: Ground truth outcome (0 or 1)
- `yes_won`: Binary label indicating if "Yes" outcome won

## Usage

```python
import pandas as pd

df = pd.read_parquet("v0.1.0/02_process/ohlc.parquet")
```