| --- |
| license: mit |
| task_categories: |
| - token-classification |
| - text-classification |
| - text-generation |
| language: |
| - en |
| pretty_name: ChessSet-Community |
| size_categories: |
| - 1K<n<10K |
| tags: |
| - ChessAI-Community |
| --- |
| # Chess Positions with Stockfish Evaluations |
|
|
| This dataset contains a collection of chess positions in Forsyth-Edwards Notation (FEN), each paired with a corresponding evaluation from the Stockfish chess engine. It is designed for use in training machine learning models for chess, analyzing chess positions, or for any application requiring a large set of evaluated positions. |
|
|
| ## Data Format |
|
|
| The data is stored in a simple CSV (Comma-Separated Values) format without a header row. Each line in the file represents a single chess position and its evaluation. |
|
|
| The format is as follows: |
| `"FEN_string",evaluation` |
|
|
| ### Fields |
|
|
| | Field | Description | |
| |-------------|-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------| |
| | `FEN_string`| A standard FEN string representing a specific board state. This includes piece placement, active color, castling availability, en passant target square, halfmove clock, and fullmove number. [Learn more about FEN](https://en.wikipedia.org/wiki/Forsyth%E2%80%93Edwards_Notation). | |
| | `evaluation`| The Stockfish engine's evaluation of the position in centipawns. <br> - A positive value (`+110`) indicates an advantage for White. <br> - A negative value (`-95`) indicates an advantage for Black. <br> - A value near zero suggests a balanced position. <br> - Mate-in-N is represented as `#N` (e.g., `#3` for mate in 3 for the side to move) or `#-N` (e.g., `#-2` if the side to move is being mated in 2). | |
|
|
| ### Example Data |
|
|
| ```csv |
| "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQkq - 0 1",+25 |
| "r1bqkbnr/pp1ppppp/2n5/2p5/4P3/5N2/PPPP1PPP/RNBQKB1R w KQkq - 2 3",+15 |
| "r4rk1/pp1n1ppp/2pbp3/q2n2B1/3PN3/3Q1N2/PPP2PPP/1K1R3R b - - 5 13",-80 |
| "4r1k1/pp1r1p1p/1qp1n1p1/4P3/3p1P2/2Q4P/PP1R2P1/3R2K1 w - - 0 29",#4 |
| ``` |
|
|
| ## How to Use |
|
|
| Here is a basic Python script to read and parse the data from a file named `chess_data.csv`. |
|
|
| ```python |
| import csv |
| |
| def load_chess_data(file_path='chess_data.csv'): |
| """ |
| Loads and prints chess positions and their evaluations from the dataset. |
| """ |
| positions = [] |
| try: |
| with open(file_path, 'r', encoding='utf-8') as f: |
| reader = csv.reader(f) |
| for row in reader: |
| if not row: |
| continue |
| fen_string = row[0] |
| evaluation = row[1] |
| positions.append({'fen': fen_string, 'eval': evaluation}) |
| print(f"FEN: {fen_string}, Evaluation: {evaluation}") |
| except FileNotFoundError: |
| print(f"Error: The file {file_path} was not found.") |
| except IndexError: |
| print(f"Error: Malformed row found in {file_path}.") |
| |
| return positions |
| |
| if __name__ == '__main__': |
| # Assuming your data is in 'chess_data.csv' |
| chess_dataset = load_chess_data() |
| if chess_dataset: |
| print(f" |
| Successfully loaded {len(chess_dataset)} positions.") |
| |
| ``` |
|
|
| ## Data Generation (Example) |
|
|
| This dataset was generated using: |
| * **Engine**: Stockfish 16 |
| * **Search**: 15 seconds per position |
| * **Source**: A curated list of positions from grandmaster games. |
|
|
| ## License |
|
|
| This dataset is released under the [MIT License](https://opensource.org/licenses/MIT). You are free to use, modify, and distribute it for any purpose. |
