```json { "table_name": "popular_matches", "table_type_tags": [ {"tag": "dimension", "confidence": 0.75}, {"tag": "event", "confidence": 0.30} ], "primary_type": "dimension", "insufficient_evidence": false, "entity": { "columns": ["date", "opponent_1", "opponent_2"], "unique_per_row": true }, "time_axis": { "columns": ["date"], "kind": "event_time" }, "reasoning": "The composite (date, opponent_1, opponent_2) is effectively unique per row — date Uniqueness 0.868, opponent_1 Uniqueness 0.750, opponent_2 Uniqueness 0.539, and their product far exceeds 152 rows. Each row describes a single boxing match occurrence with its attributes (punch power estimates, rounds boxed, verdict). The date is an attribute of the match, not an axis ordering actions of a repeating entity. A sibling 'fighters' table likely references opponent names, making this the catalog of match occurrences those fighters participated in → dimension. The occurrence-like nature (matches happened) gives a weak secondary event reading, but no repeating entity exists whose actions the rows order." } ```