chorcat/rukh-elo-bins
chorcat/rukh-elo-bins A balanced sample of rukh-games-1800: up to n_per_bin games per 100-Elo bin of the average rating of both players, for Elo conditioning experiments. Part of Rukh, a chess language model built from scratch as a course on generative and agentic AI. Every derived dataset ships with the exact filters and counts of its manifest.json, so it can be regenerated with rukh data elo-bins. Files File Bytes SHA-256 games.parquet 65955015… See the full description on the dataset page: https://huggingface.co/datasets/chorcat/rukh-elo-bins.
chorcat/rukh-elo-bins
A balanced sample of rukh-games-1800: up to n_per_bin games per 100-Elo bin of the average rating of both players, for Elo conditioning experiments.
Part of Rukh, a chess language model built from scratch as a course on generative and agentic AI. Every derived dataset ships with the exact filters and counts of its manifest.json, so it can be regenerated with rukh data elo-bins.
Files
Rows: 423829.
Columns
Provenance and filters
Source: rukh-games-1800 (from Lichess/standard-chess-games) (months 2025-01, 2025-02).
{
"bin": "((white_elo + black_elo) // 2) // 100 * 100",
"n_per_bin": 50000,
"seed": 42
}Counts:
{
"1800": 50000,
"1900": 50000,
"2000": 50000,
"2100": 50000,
"2200": 50000,
"2300": 50000,
"2400": 50000,
"2500": 44982,
"2600": 17522,
"2700": 5669,
"2800": 2316,
"2900": 1436,
"3000": 1802,
"3100": 102
}Generated on 2026-09-19 with rukh 0.0.1.
In the course
- Built in M1 · Del PGN al tensor: los labs, the lesson that runs every command behind this repository.
- Measured in the single results table: every stage of the course, the same suite, the same day.
- Bring it to the paths the configs read:
uv run rukh pull rukh-elo-bins.
License
CC0 1.0. The games, puzzles and evaluations come from the Lichess open database (https://database.lichess.org, https://huggingface.co/Lichess), published under CC0; the derived files keep that license. Please credit Lichess when you use them.
