Lichess Chess Position Evaluations (Stockfish)
- Provider
- Lichess (lichess.org)
- License
- CC0-1.0 (source)
- Size
- 22 files
- Format
- Parquet (in sampled files)
- Stored at
- Hugging Face (https://huggingface.co/datasets/Lichess/chess-position-evaluations)
- Access
- Free and public. No account with the data provider is needed.
- Category
- Machine Learning & AI
- Source
- https://database.lichess.org/#evals
About Lichess Chess Position Evaluations (Stockfish)
Chess positions evaluated with the Stockfish engine, produced by and for the Lichess analysis board, where various builds of Stockfish run inside users' browsers. The card reports 416,442,401 positions and monthly updates. Each row gives a position as a FEN string, the principal variation as a sequence of UCI moves, the search depth, the number of kilonodes searched and the score either as centipawns (cp) or as a forced mate distance (mate); a position can appear several times with different depths and lines. Listing data/ (measured 2026-11-14) shows 20 Parquet shards of about 2.2 GB each, and the repository uses about 609 GB. The data serve as supervised targets for training neural chess evaluation functions and value networks, for distillation of engine knowledge into small models, for benchmarking evaluation accuracy and for studying position complexity. Load it with Hugging Face datasets (load_dataset with split train, or streaming=True), pyarrow, polars or DuckDB, and parse positions with python-chess. Lichess releases its database exports under CC0.
How to load Lichess Chess Position Evaluations (Stockfish) in Python
from datasets import load_dataset
ds = load_dataset("Lichess/chess-position-evaluations")
print(ds)Data structure
File types in sampled files: Parquet
Files
Example file: data/data_0000.parquet
Analyze this dataset without downloading it
On Scigantic, Lichess Chess Position Evaluations (Stockfish) 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.