**Yes, but the exact duplicates are negligible. The near-duplicates are more of an issue.** **Exact duplicates (every column identical)** - `raw_kaggle` has 237,205 rows but 237,203 distinct, so 2 rows are exact repeats. - `return_kaggle` has 237,196 rows and 237,194 distinct, so 2 rows are repeats. - `serve_kaggle` has 237,185 rows and 237,183 distinct, so 2 rows are repeats. - `players_tournament_man_` (6,422 rows) and `players_man_` (462 rows) have no exact duplicates. **Near-duplicates (same match key, different values)** - I checked the three match-level tables against their inferred grain of one row per player per match: Name, Date, Tournament and Round. - `raw_kaggle` has 1,250 extra rows on that key, `return_kaggle` has 1,251, and `serve_kaggle` has 1,250. That is about 0.5% of each table. - Only 2 of those extra rows per table are exact copies. The other roughly 1,248 share a key but differ in at least one column. - `players_tournament_man_` has 0 extra rows on its key (name, year, tournament), so it is clean. **Interpretation** - The three match tables have nearly the same counts of key collisions (1,250, 1,251, 1,250), so the collisions likely come from the same underlying matches appearing in all three. - I did not inspect the colliding rows. Possible causes include the same player meeting the same opponent twice in one round and tournament, for example a replay or qualifying/main-draw overlap, or a key that is too coarse. Adding the opponent or another column may separate them. They may also be true duplicates with slightly different values. **Recommendation** - Drop the 2 exact duplicate rows in each table. - Inspect a sample of the roughly 1,250 key collisions before deduplicating on the key, since removing them blindly could delete real matches.