alphalab-research/polymarket-2024

该数据集名为Polymarket 2024 — NegRisk Wash-Flagged Trade Dataset (v3.1)，是一个用于表格分类和时间序列预测任务的数据集。它覆盖了2024年1月5日至2024年12月7日的时间窗口，包含约11.1个月的数据，涉及16,959,221笔交易，总名义金额为3,545,492,028美元，覆盖199个独特的Polymarket V1 NegRisk市场。数据源来自Polymarket V1 CTFExchange通过Goldsky子图回填，并过滤到WS-N家族重建宇宙（即38个最高优先级的NegRisk市场）。数据集使用CTFPositionDeriver（Alpha Lab WS-BF.2）将78位十进制market_id（ERC-1155 CTF代币ID）桥接到32字节NegRisk family_id。每行数据包含wash_flag，这是V3WashDetectorWithThetaTuning的输出，基于wash_score和每市场θ阈值进行标记，用于检测洗售交易。数据集还包括交易ID、时间戳、钱包地址、交易方向、大小、哈希值、家庭ID、问题ID、洗售分数、标志、原因和方法版本等列。数据按年份和月份分区存储为Parquet文件。数据集遵循CC-BY-NC-4.0许可证，仅用于研究或非商业用途。The dataset is named Polymarket 2024 — NegRisk Wash-Flagged Trade Dataset (v3.1), designed for tabular classification and time-series forecasting tasks. It covers the period from January 5, 2024, to December 7, 2024, spanning approximately 11.1 months, with 16,959,221 trades, a total notional value of $3,545,492,028, and 199 unique Polymarket V1 NegRisk markets. The data is sourced from Polymarket V1 CTFExchange via the Goldsky subgraph backfill, filtered to the WS-N family-reconstruction universe (the 38 highest-priority NegRisk markets). The dataset uses CTFPositionDeriver (Alpha Lab WS-BF.2) to bridge the 78-digit decimal market_id (the ERC-1155 CTF token id) to a 32-byte NegRisk family_id. Each row includes a wash_flag, which is the output of the V3WashDetectorWithThetaTuning, based on a wash_score and a per-market θ threshold for detecting wash trading. The schema includes columns such as trade_id, trade_at_utc, proxy_wallet, counterparty_wallet, side, size, transaction_hash, order_hash, family_id, family_id_unknown, question_id, question_index, wash_score, wash_flag, theta_used, wash_reasoning, wash_method_version, archiver_run_id, and fetched_at_utc. The data is partitioned by year and month into Parquet files. The dataset is licensed under CC-BY-NC-4.0 for research and non-commercial use.