{
    "table_name": "serve_kaggle",
    "primary_key": "Name, Date, Tournment, Rd",
    "candidates": [
        {
            "rank": 0,
            "partition_key": "Name",
            "partition_key_label": "E",
            "partition_key_reasoning": "Name is the player whose match-serve statistics are recorded. It has 462 unique values with RPUV ~513 and top-3 coverage of only 1.74%, so no single player dominates. RPUV is above the ideal ~100 range but still gives evenly sized partitions, and Name is the natural organizer of a per-player timeline.",
            "clustering_key": "Date",
            "clustering_key_label": "S",
            "clustering_key_reasoning": "Date is the match start date and orders each player's matches over time. It is stored as text in d-Mon-yyyy format (e.g. 2‑Oct‑2024), so it must be parsed to a date before sorting because lexical order is wrong. It has 1424 unique values and 0% nulls. Within a tournament, Rd can break ties.",
            "world_line": "Name",
            "session_column": "Tournment",
            "confidence": "MEDIUM",
            "notes": "Each row is one player's serve stats for one match, so the table is a per-player event log. The composite key Name+Date+Tournment+Rd identifies a row, since a player plays at most one match per round of a tournament. The 'against' column is a messy, fragmented text field that is almost unique (231,760 distinct values), so it is unusable as a key. Its weakness is that it embeds the opponent with mangled formatting. Tournment (3572 values, RPUV ~66) could group matches into tournament sessions."
        },
        {
            "rank": 1,
            "partition_key": "Tournment",
            "partition_key_label": "E",
            "partition_key_reasoning": "Tournment has 3572 unique values with RPUV ~66, inside the preferred 2–100 range, and top-3 coverage of 5.01%. It works as an alternative perspective that groups all player matches within a tournament. Values include the year or edition as separate entries for some events.",
            "clustering_key": "Date",
            "clustering_key_label": "S",
            "clustering_key_reasoning": "Date orders matches chronologically within a tournament, though it must be parsed from text. Rd can be used as a secondary ordering.",
            "world_line": "Tournment",
            "session_column": null,
            "confidence": "LOW",
            "notes": "Alternative tournament-centric view for analyzing the draw progression. It is less natural than the player view because the tournament name is reused across years, so the same name covers many separate editions."
        }
    ]
}