```json { "table_name": "ittf_rankings", "primary_key": "ID,YearNum,MonthNum,WeekNum", "candidates": [ { "rank": 0, "partition_key": "YearNum", "partition_key_label": "T", "partition_key_reasoning": "YearNum is the primary temporal dimension with RPUV=13,563.55 (20 unique years). This provides complete snapshots of all player rankings at each time period. Partitioning by year gives manageable file sizes and aligns with the state table pattern where we want all entities' states at a given time together.", "clustering_key": "ID", "clustering_key_label": "X", "clustering_key_reasoning": "ID is the entity (player) being measured, with RPUV=40.99 (6,618 unique players). Clustering by ID within each year's partition groups all temporal snapshots for each player together, enabling efficient queries for player ranking history. Monotonicity=0.504 shows reasonable ordering.", "world_line": null, "session_column": null, "confidence": "HIGH", "notes": "This is a state table tracking player rankings over time. Each (ID, YearNum, MonthNum, WeekNum) combination represents a player's ranking snapshot at a specific point in time. Partitioning by YearNum (T) gives complete cross-sections of all players at each time period, which is the canonical state table pattern. MonthNum and WeekNum provide finer temporal granularity within each year but are included in the composite primary key rather than used for partitioning." }, { "rank": 1, "partition_key": "ID", "partition_key_label": "E", "partition_key_reasoning": "ID (player) with RPUV=40.99 represents an alternative perspective viewing this as a path table tracking each player's ranking journey over time. Each player has ~41 ranking snapshots on average across the dataset.", "clustering_key": "YearNum", "clustering_key_label": "S", "clustering_key_reasoning": "YearNum provides temporal ordering with monotonicity=0.000 (not strictly increasing but sequential). Combined with MonthNum and WeekNum, this creates the timeline for each player's ranking evolution.", "world_line": "ID", "session_column": null, "confidence": "MEDIUM", "notes": "Alternative path-oriented view: partition by player (E=ID) and sort by time (S=YearNum). This perspective optimizes for queries like 'show me player X's complete ranking history' rather than 'show me all rankings in year Y'. The classifier tagged this as state=0.85, path=0.25, so this path perspective is valid but secondary. Would require composite clustering (YearNum, MonthNum, WeekNum) for proper temporal ordering." } ] } ```