NBA_Games

<p align="left"> <a href="README_CN.md">中文</a>&nbsp;｜&nbsp;English </p> <br> # NBA Full-Game Video Dataset <p align="center"> <img src="images/nba.jpeg" alt="NBA Full-Game Video Metadata Dataset" width="900"/> </p> <p align="center"> <a href="https://github.com/choucisan/nba_games"><img src="https://img.shields.io/badge/GitHub-NBA_Dataset-181717?style=for-the-badge&logo=github" alt="GitHub"></a> &nbsp; <a href="https://huggingface.co/datasets/choucsan/NBA_Games"><img src="https://img.shields.io/badge/%F0%9F%A4%97_HuggingFace-Dataset-yellow?style=for-the-badge" alt="Hugging Face"></a> &nbsp; <a href="https://modelscope.cn/datasets/choucisan/NBA_Games"><img src="https://img.shields.io/badge/ModelScope-Dataset-624aff?style=for-the-badge" alt="ModelScope"></a> &nbsp; <a href="https://choucisan.github.io/collections/nba_games"><img src="https://img.shields.io/badge/Blog-Post-blue?style=for-the-badge" alt="Blog"></a> &nbsp; <a href="https://www.xiaohongshu.com/explore/6a0bfa950000000036019e44?xsec_token=ABMOfUoZI4fqfkoQyBURZvY1X7Kz11uX9Z4-15k_6UueE=&xsec_source=pc_user"><img src="https://img.shields.io/badge/RedNote-Post-red?style=for-the-badge" alt="RedNote"></a> &nbsp; <a href="https://choosealicense.com/licenses/mit"><img src="https://img.shields.io/badge/License-MIT-green?style=for-the-badge" alt="License"></a> </p> This dataset provides metadata, official statistics, and official play-by-play annotations for full-length NBA game videos available on YouTube. Instead of redistributing video files, we provide YouTube video IDs and URLs so users can download videos independently when their use case and local policies allow it. The dataset links long-form basketball videos with structured NBA.com game data. Each retained game has a verified YouTube video reference, a cleaned matchup and date, an official NBA box score, and, when available, official play-by-play event sequences. --- ## Pipeline The dataset is built through a multi-stage cleaning and verification pipeline: 1. **Playlist extraction**: We started from the well-known YouTube NBA full-game playlist [`PLNbBj4TorBWerlzB1A5iM3XjwigqFG8sY`](https://www.youtube.com/playlist?list=PLNbBj4TorBWerlzB1A5iM3XjwigqFG8sY), containing hundreds of historic and classic games. 2. **Valid-video filtering**: From 595 playlist entries, we retained 217 valid full-game candidates with real titles and durations. 3. **Metadata normalization**: Cleaned titles into canonical matchup format (e.g. `Golden State Warriors vs. Oklahoma City Thunder`). 4. **External match linking**: Linked videos to official NBA.com games via date and team name matching. 5. **Box score and play-by-play retrieval**: Queried the official NBA stats API for box scores and play-by-play event sequences. --- ## Dataset Structure ``` NBA_Games/ ├── images/ │ └── nba.jpeg ├── nba_games.jsonl # 189 games metadata ├── nba_games_box_score.jsonl # Official box scores ├── nba_games_play_by_play.jsonl # Play-by-play event sequences └── pages/ # Individual game detail pages ``` ## Dataset Overview - **Total Games**: 189 (verified full-game videos) - **Data Format**: JSONL | Field | Type | Description | |-------|------|-------------| | `id` | string | YouTube video ID | | `url` | string | YouTube video URL | | `title` | string | Game title with matchup info | | `date` | string | Game date | | `duration` | int | Video duration in seconds | | `description` | string | Video description from YouTube | ## Usage ```python from datasets import load_dataset dataset = load_dataset("choucsan/NBA_Games") # or from ModelScope from modelscope import MsDataset dataset = MsDataset.load("choucisan/NBA_Games") ``` ## License MIT License. ## Contact - 📧 **[e-mail]** - 🤗 [Hugging Face](https://huggingface.co/datasets/choucsan/NBA_Games) - 🐱 [ModelScope](https://modelscope.cn/datasets/choucisan/NBA_Games)