Dark Mode
AlphaZero vs Stockfish 110 Games Data
Product Reviews & Feedback
Tags and Keywords
Trusted By
"No reviews yet"
Free
About
Offers 110 historical chess games played between the elite artificial intelligence programmes, AlphaZero and Stockfish 8. The dataset captures the full moves and results from these high-stakes computer matches, which originated from a raw PGN file provided by Deepmind. The data has been converted into a structured CSV format, making it readily available for detailed analysis of AI versus AI strategy, performance, and game outcomes.
Columns
- row number: Sequential index for the record.
- game_id: Unique identifier for each game, generated during the data processing.
- game_order: The order in which the game appeared in the original PGN source file.
- event: The name of the tournament or match, recorded as "Computer Match".
- site: The physical location where the match was played, recorded as "London, UK".
- date_played: The specific calendar date on which all games were conducted, 18 January 2018.
- round: An internal identifier related to the match round.
- white: Identifies the player using the White pieces (AlphaZero or Stockfish 8).
- black: Identifies the player using the Black pieces (Stockfish 8 or AlphaZero).
- result: The official outcome of the game (e.g., 1/2-1/2, 1-0).
- winner: Identifies the declared winner of the match (draw, AlphaZero, or Stockfish 8).
- loser: Identifies the declared loser of the match (draw, Stockfish 8, or AlphaZero).
- date_created: The date the game record was processed and added to the present file.
- file_name: The name of the raw source file,
alphazero_vs_stockfish_all.pgn.
Distribution
The dataset contains 110 individual records, corresponding to the 110 games played between the engines. It is structured with 14 distinct columns. The data is supplied as a CSV file (
alphazero_stockfish_games_df.csv) and is approximately 22.24 kB in size. The data quality is excellent, with 100% valid records and no missing or mismatched values across the entries.Usage
The data is ideal for analysing high-level chess AI strategy, focusing on Deepmind’s AlphaZero against the established engine Stockfish 8. It can be used to study engine opening preferences, evaluate the high proportion of draw results (65% of games), and investigate how often AlphaZero prevailed (winning 32% of the matches). It serves as a foundational dataset for machine learning models seeking to replicate or understand self-play reinforcement learning outcomes.
Coverage
The geographic scope is limited to the match location: London, UK. The temporal coverage is a single day of play, 18 January 2018. The dataset includes 110 games, with all starting from the initial board position. The players covered are the AI engines AlphaZero and Stockfish 8. AlphaZero played White in 73% of the recorded games.
License
CC0: Public Domain
Who Can Use It
- Artificial Intelligence Researchers: For training models or comparing different AI approaches (e.g., Monte Carlo Tree Search vs. traditional minimax search).
- Chess Programmers: To test and benchmark proprietary chess algorithms against historical elite engine matches.
- Data Scientists: For statistical analysis of game outcomes, move sequencing, and performance metrics in closed systems.
Dataset Name Suggestions
- AlphaZero vs Stockfish 110 Games Data
- AI Chess Engine Matchup (2018)
- Deepmind Chess Match Records
- Stockfish-AlphaZero Game History
Attributes
Original Data Source: AlphaZero vs Stockfish 110 Games Data
QUALITYLoading...