smf-ulm/polymarket-quant-bench

Polymarket Quant Bench是一个教学级衍生数据集，基于Polymarket（Polygon区块链上的去中心化预测市场）的链上交易历史，专为指标工程和向前回测任务设计。该版本将上游数据集的每笔交易填充流替换为每小时和每日分辨率的每代币OHLCV柱状图，并限制于高流动性已解决市场，确保每个市场都有已知结果用于盈亏评分，并有足够深度用于滚动指标计算。数据集包含市场元数据（36,831行，每个市场一行）、小时OHLCV柱状图（12,655,266行，每个代币每小时一行）和日度OHLCV柱状图（1,462,282行，每个代币每日一行），所有数据均以Parquet格式分片存储。Polymarket Quant Bench is a pedagogical derivative dataset based on the on-chain transaction history of Polymarket, a decentralized prediction market built on the Polygon blockchain, designed specifically for indicator engineering and forward backtesting tasks. This version replaces the raw trade fill flow of the upstream dataset with per-token OHLCV candlesticks at hourly and daily resolutions. It is restricted to highly liquid resolved markets, ensuring that each market has a known outcome for profit and loss (P&L) scoring and sufficient depth for rolling indicator calculations. The dataset includes market metadata (36,831 rows, one row per market), hourly OHLCV candlesticks (12,655,266 rows, one row per token per hour), and daily OHLCV candlesticks (1,462,282 rows, one row per token per day), all stored in sharded Parquet format.