CoolFace
Datasetpublic

Lichess/standard-chess-games

[!CAUTION] This dataset is still a work in progress and some breaking changes might occur. Lichess Rated Standard Chess Games Dataset Dataset Description 6,771,826,271 standard rated games, played on lichess.org, updated monthly from the database dumps. This version of the data is meant for data analysis. If you need PGN files you can find those here. That said, once you have a subset of interest, it is trivial to convert it back to PGN as shown in the Dataset… See the full description on the dataset page: https://huggingface.co/datasets/Lichess/standard-chess-games.

sourceHugging Facecc0-1.0updated 11mo agoView on Hugging Face
74likes617kdownloads
README.md93 linesDownload Raw Back to root
1---2license: cc0-1.03pretty_name: Lichess Standard Rated Games4configs:5- config_name: default6  data_files:7  - split: train8    path: data/**/train-*9tags:10- chess11- games12- game13- lichess14- tabular15size_categories:16- 1B<n<10B17---18 19> [!CAUTION]20> This dataset is still a work in progress and some breaking changes might occur.21>22 23# Lichess Rated Standard Chess Games Dataset24 25## Dataset Description26 27**6,771,826,271** standard rated games, played on [lichess.org](https://lichess.org), updated monthly from the [database dumps](https://database.lichess.org/#standard_games).28 29This version of the data is meant for data analysis. If you need PGN files you can find those [here](https://database.lichess.org/#standard_games). That said, once you have a subset of interest, it is trivial to convert it back to PGN as shown in the [Dataset Usage](#dataset-usage) section.30 31This dataset is hive-partitioned into multiple parquet files on two keys: `year` and `month`:32 33```bash34.35├── data36│   └── year=201537│       ├── month=0138│       │   ├── train-00000-of-00003.parquet39│       │   ├── train-00001-of-00003.parquet40│       │   └── train-00002-of-00003.parquet41│       ├── month=0242│       │   ├── train-00000-of-00003.parquet43│       │   ├── train-00001-of-00003.parquet44│       │   └── train-00002-of-00003.parquet45│       ├── ...46```47 48### Dataset Usage49 50<!-- Using the `datasets` library:51 52```python53from datasets import load_dataset54dset = load_dataset("Lichess/chess-evaluations", split="train")55```56Using the `polars` library:57 58Using DuckDB:59 60Using `python-chess`: -->61 62 63## Dataset Details64 65### Dataset Sample66 67<!-- One row of the dataset looks like this:68 69```python70{71  "Event":,72  "Site":,73}74``` -->75 76### Dataset Fields77 78<!-- Every row of the dataset contains the following fields:79 80- **`Event`**: `string`, 81- **`Site`**: `string`, -->82 83### Notes84 85- About 6% of the games include Stockfish analysis evaluations: [%eval 2.35] (235 centipawn advantage), [%eval #-4] (getting mated in 4), always from White's point of view.86 87- The WhiteElo and BlackElo tags contain Glicko2 ratings.88 89- The `movetext` column contains clock information as PGN %clk comments since April 2017.90 91- The schema doesn't include the `Date` header, typically part of the [Seven Tag Roster](https://en.wikipedia.org/wiki/Portable_Game_Notation#Seven_Tag_Roster) as we deemed the `UTCDate` field to be enough.92 93- A future version of the data will include the addition of a `UCI` column containing the corresponding moves in [UCI format](https://en.wikipedia.org/wiki/Universal_Chess_Interface).
Lichess/standard-chess-games · CoolFace