Sign in to view source links and access this dataset
Description
Syzygy soft MultiPV endgame dataset provides perfect ≤5-piece chess positions labeled via Syzygy WDL (soft top-8 moves). Created by user 'avewright', it is compatible with ChessTransformer soft-cache training and was last updated on July 15, 2026. The dataset includes columns for board state, evaluation, and soft policy targets.
Use Cases
Training reinforcement learning agents for chess endgames based on perfect tablebase evaluations.
Benchmarking chess AI policy networks using the soft policy targets (soft_indices[8], soft_probs[8]).
Analyzing endgame positions and strategies using the provided board state, WDL, and DTZ columns.
Strengths
Positions are labeled with perfect ≤5-piece endgame evaluations from Syzygy tablebases.
Includes soft policy targets for the top-8 moves, designed for compatibility with ChessTransformer training.
Dataset was updated on July 15, 2026, suggesting recent maintenance.
Limitations
Description metadata is limited; actual data quality requires manual inspection after download.
Row count is unknown, which may limit suitability assessment.
Column-level documentation is absent; field semantics must be inferred after download.
Provenance
Source
huggingface
Collection Method
Derived from Syzygy endgame tablebases.
Freshness
Last updated 2026-07-15 02:03:59.
License is unknown; users should verify licensing terms before use.