cdechoch/nba-data-archive

NBA数据档案是一个Parquet格式的数据集，镜像了NBA从1996/97赛季到2025/26赛季的play-by-play（逐场比赛数据）、shot detail（投篮细节）和player-matchup（球员对位）数据。该数据集作为HoopsMatic.com分析工具和HoopsHype编辑自动化的数据层构建和维护，并公开供其他研究人员和爱好者使用，以避免重复构建相同的数据管道。数据集提供两种互补的布局：per_season/目录按赛季和赛季类型（常规赛或季后赛）组织数据，适用于需要特定时间段数据的场景；merged/目录将所有赛季和赛季类型的数据合并为单个文件，适用于需要历史全貌分析的场景。数据包括多种类型，如shotdetail（涵盖每个投篮的球场坐标、距离、命中/未命中等信息）、matchups（球员对位数据）、nbastats（经典play-by-play数据）等。数据集基于Apache License 2.0许可，原始数据来源于stats.nba.com、data.nba.com等上游渠道，并以CSV和Parquet格式提供。The NBA Data Archive is a Parquet mirror of NBA play-by-play, shot detail, and player-matchup data from the 1996/97 season through 2025/26. It is built and maintained as the data layer for HoopsMatic.com analytics tools and HoopsHype editorial automation, and is made public so other researchers and tinkerers do not have to rebuild the same pipeline. The dataset offers two complementary layouts: the per_season/ directory organizes data by season and season type (regular or playoffs), suitable for scenarios requiring data from specific time periods; the merged/ directory concatenates all seasons and season types into single files, ideal for comprehensive historical analysis. Data types include shotdetail (covering every shot with court coordinates, distance, make/miss, etc.), matchups (player-vs-player matchup data), nbastats (classic play-by-play data), and more. It is licensed under Apache License 2.0, with original data sourced from upstream channels such as stats.nba.com and data.nba.com, and provided in both CSV and Parquet formats.