Lichess Chess Puzzles Database
- Provider
- Lichess (lichess.org)
- License
- CC0-1.0 (source)
- Size
- 5 files
- Format
- Parquet (in sampled files)
- Stored at
- Hugging Face (https://huggingface.co/datasets/Lichess/chess-puzzles)
- Access
- Free and public. No account with the data provider is needed.
- Category
- Machine Learning & AI
- Source
- https://database.lichess.org/#puzzles
About Lichess Chess Puzzles Database
Tactical chess puzzles generated from real Lichess games, rated and tagged by theme. The card reports 6,157,341 puzzles, updated monthly. Generating the initial set took more than 50 years of CPU time: Lichess went through 300 million analysed games and re-analysed interesting positions with Stockfish 12 to 15 NNUE at 40 meganodes, then players solving the puzzles on lichess.org produced Glicko-2 ratings for them. Each row holds a puzzle id, the source game id and move, the starting position as FEN, the solution as UCI moves, the puzzle rating and rating deviation, popularity, number of plays, theme tags (fork, pin, mate in 2, endgame and so on) and opening tags. Listing data/ (measured 2026-11-14) shows three Parquet shards of about 0.3 GB. Uses include tactics training, measuring the reasoning ability of chess engines and language models, puzzle difficulty prediction, curriculum learning and studies of human problem solving. Load it with Hugging Face datasets, pandas or pyarrow and replay solutions with python-chess. Lichess releases its database exports under CC0.
How to load Lichess Chess Puzzles Database in Python
from datasets import load_dataset
ds = load_dataset("Lichess/chess-puzzles")
print(ds)Data structure
File types in sampled files: Parquet
Files
Example file: data/train-00000-of-00003.parquet
Analyze this dataset without downloading it
On Scigantic, Lichess Chess Puzzles Database opens in a hosted JupyterLab notebook with the data already mounted, so you can read files where they are stored instead of copying them first. Each notebook includes an AI assistant that can see the mounted files and write and run analysis code. Open the dataset page to launch a notebook, or browse the full catalog.
This page is also available as Markdown and as JSON.