Zweig Chess Engine Processed: Human-Aligned Chess Games from Lichess 2025

11.15 million processed chess games from the Lichess Open Database for the year 2025. The dataset was curated by ygkla and is designed to train human-aligned chess models. It was last updated on February 2, 2026.

Use Cases

• Training chess AI to predict human moves based on the dataset's focus on human-aligned play.
• Developing ELO-conditioned neural networks based on the dataset's stated design purpose.
• Analyzing human chess behavior and decision-making patterns from the large collection of Lichess games.

Strengths

• Contains 11.15 million chess games, providing a substantial base for model training.
• Specifically curated from the Lichess Open Database for the year 2025, offering a defined temporal scope.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

Lichess Open Database

Collection Method

Curated and processed from the open database.

Time Range

Year 2025

Freshness

Last updated 2026-02-02 20:06:57; freshness should be verified.