The Art of Thinking in 64 Squares : Chess Dataset

This dataset consists of over 30,000 real chess games that have been downloaded from Chess.com. Data were collected ethically from authors' accounts, authors' friends' data, and even Grandmaster players' data whose game history they donated for research purposes. This dataset includes public chess games played on Chess.com, including titled players' games. Move data (PGN) alone is provided; proprietary annotations or copyrighted content from Chess.com are not used. The data set contains players from beginners to Grandmasters (GMs) and contains complete information for all games. Elo rating, game result (win/loss/draw), moves played, opening names, game type (Blitz, Bullet, Rapid), and game termination (e.g., Checkmate, Timeout, or Resignation) are included. Other columns helpful for analysis were added, e.g., the difference in Elo between both players and acronyms of opening names. Data is also separated into various sheets based on Elo range, game type, and player identity to make it easier to analyze. This data set is a solid base to train AI chess models with reinforcement learning, study player behavior, test chess coaching systems, and build learning tools. It is in Excel (.xlsx) format to be easily used with many applications like Power BI and Python.