```json { "table_name": "2019_WILD CARD WEEKEND_plays", "primary_key": "Season,Week,AwayTeam,HomeTeam,DriveNumber,PlayNumberInDrive", "candidates": [ { "rank": 0, "partition_key": "TeamWithPossession", "partition_key_label": "E", "partition_key_reasoning": "TeamWithPossession is the entity whose play-by-play timeline we trace. RPUV=85.62 (8 unique teams, 685 rows) is excellent for partitioning. Top 3 coverage is 42.63%, well below 50%. Semantically, this groups all plays by the team executing them, enabling team-centric analysis of offensive sequences.", "clustering_key": "PlayNumberInDrive", "clustering_key_label": "S", "clustering_key_reasoning": "PlayNumberInDrive is the sequence axis ordering plays within each drive. Monotonicity=0.874 confirms strong sequential ordering. This column orders plays 1→17 within each possession, creating chronological play sequences for each team's drives.", "world_line": "TeamWithPossession", "session_column": "DriveNumber", "confidence": "HIGH", "notes": "Primary perspective: partition by team to analyze offensive play sequences. DriveNumber serves as session_column to group plays into possession units. The composite primary key (Season,Week,AwayTeam,HomeTeam,DriveNumber,PlayNumberInDrive) uniquely identifies each play in the game." }, { "rank": 1, "partition_key": "derived:concat(AwayTeam,'_vs_',HomeTeam)", "partition_key_label": "E", "partition_key_reasoning": "Game-level perspective: partition by matchup (AwayTeam+HomeTeam combination). This creates 4 unique game partitions from the 4 matchups in Wild Card Weekend. RPUV would be ~171 (4 games, 685 plays). Enables game-centric analysis where all plays from a single contest are co-located.", "clustering_key": "PlayNumberInDrive", "clustering_key_label": "S", "clustering_key_reasoning": "Same sequential ordering within drives. Combined with DriveNumber and TeamWithPossession in queries, this reconstructs the full game timeline play-by-play.", "world_line": "derived:concat(AwayTeam,'_vs_',HomeTeam)", "session_column": "DriveNumber", "confidence": "MEDIUM", "notes": "Alternative game-level perspective: useful for analyzing entire games as units rather than individual team performances. Requires derived column for partition key. Less natural than team-based partitioning but valid for game-focused analytics." } ] } ```