README.md

---
license: cc0-1.0
size_categories:
- 1K<n<10K
pretty_name: Chess Openings
dataset_info:
  features:
  - name: eco-volume
    dtype: string
  - name: eco
    dtype: string
  - name: name
    dtype: string
  - name: img
    dtype: image
  - name: pgn
    dtype: string
  - name: uci
    dtype: string
  - name: epd
    dtype: string
  splits:
  - name: train
    num_bytes: 215781488.706
    num_examples: 3522
  download_size: 215237282
  dataset_size: 215781488.706
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
tags:
- chess
- lichess
- games
- game
---

# Dataset Card for Lichess Puzzles

<!-- Provide a quick summary of the dataset. -->
## Dataset Description

3522 chess openings with their [Encyclopaedia of Chess Openings (ECO)](https://en.wikipedia.org/wiki/Encyclopaedia_of_Chess_Openings) classification.  

### Dataset Creation
Creating this dataset is detailed in its original [GitHub repository](https://github.com/lichess-org/chess-openings).

Updates to the original repo will also be reflected in this version. Dataset last updated on June 10th, 2025.

### Dataset Usage

Using the `datasets` library:

```python
from datasets import load_dataset

dset = load_dataset("Lichess/chess-openings", split="train")
```
 Using the `pandas` library:

 ```python
import pandas as pd

df = pd.read_parquet("hf://datasets/Lichess/chess-openings/data/train-00000-of-00001.parquet")
```

Using the `polars` library:

```python
import polars as pl

df = pl.read_parquet('hf://datasets/Lichess/chess-openings/data/train-00000-of-00001.parquet')
```

## Dataset Details

### Dataset Sample

One row of the dataset looks like this:

```python
{
 'eco-volume': 'B',
 'eco': 'B31',
 'name': 'Sicilian Defense: Nyezhmetdinov-Rossolimo Attack, Fianchetto Variation, Totsky Attack',
 'pgn': '1. e4 c5 2. Nf3 Nc6 3. Bb5 g6 4. O-O Bg7 5. c3 Nf6 6. Qa4',
 'uci': 'e2e4 c7c5 g1f3 b8c6 f1b5 g7g6 e1g1 f8g7 c2c3 g8f6 d1a4',
 'epd': 'r1bqk2r/pp1pppbp/2n2np1/1Bp5/Q3P3/2P2N2/PP1P1PPP/RNB2RK1 b kq -'}
}
```
### Dataset Fields

Every row of the dataset contains the following fields:

- **`eco-volume`**: `string`, the ECO parent category. Can be either `A`, `B`, `C`, `D`, or `E`.
- **`eco`**: `string`, the ECO.
- **`name`**: `string`, the name of the opening in English.
- **`pgn`**: `int`, the sequence of moves leading to the opening position. If not unique, the moves are the most common ones leading to that opening.
- **`uci`**: `int`, the same moves described by the `pgn` field in [UCI notation](https://backscattering.de/chess/uci/#move).
- **`epd`**: `int`, the [EPD](https://www.chessprogramming.org/Extended_Position_Description) ([FEN](https://en.wikipedia.org/wiki/Forsyth%E2%80%93Edwards_Notation) without move numbers) of the opening position, en passant field only if legal

### Dataset Conventions

* Title case is used for opening names.
* Names are structured like `Opening family: Variation, Subvariation, ...`,
  e.g., `Sicilian Defense: Najdorf Variation, English Attack`.
* The suggested way to classify games is to play moves backwards until
  a named position is found. To make this work well with common transpositions,
  multiple entries for a single opening may be added.
* However, each name has a unique *shortest* line. If necessary,
  a distinguishing move is appended, e.g.,
  `King's Gambit Accepted: Schurig Gambit, with Bb5`.

## Additional Information

- To contribute to the dataset, send a PR on the [dataset's GitHub repository](https://github.com/lichess-org/chess-openings/blob/master/CONTRIBUTING.md).