Across the dataset, 18 columns are profiled as categorical. The counts below come from the profiler's approximate distinct counts (step 2), so treat them as estimates, not exact values. **Low cardinality (1 to 24 values)** - `return_kaggle.against` has about 1 value, so it is effectively constant. - `Surface` has about 4 values in each of `raw_kaggle`, `return_kaggle` and `serve_kaggle`. These are the court surfaces. - `Rd` (round) has about 14 values in each of the same three tables. - `players_tournament_man_.year` has about 24 values. **Medium cardinality (29 to 236 values)** - `raw_kaggle.DFs` (double faults) has about 29 values. - `raw_kaggle.Aces` has about 54 values. - `raw_kaggle.vA` has about 62 values. - `raw_kaggle.2SP` has about 134 values. - `raw_kaggle.1SP` has about 189 values. - `serve_kaggle.Df%` has about 231 values. - `serve_kaggle.Bpsvd` (break points saved) has about 236 values. **Higher cardinality (491 values)** - `Name` has about 491 values in each of `raw_kaggle`, `return_kaggle` and `serve_kaggle`, meaning about 491 distinct players. The `players_man_` table has 462 rows, so the approximate 491 is probably a slight overestimate or reflects names that differ between tables. **Caveats** - The profiler labelled a column "categorical" mainly when its distinct count was low. The numeric measures above (Aces, DFs, 1SP, 2SP, vA) are really numbers that happen to have few distinct values. They are not true categories. - Other columns are stored as text and classed as "text" rather than categorical, so they are not counted here. Examples are `Tournament` (about 3,963 values), `Score` (about 10,386) and `against` in `raw_kaggle` and `serve_kaggle` (about 252,129). - Most of the return and serve percentage fields, such as `RPW` and `1st%`, are text columns with 200 to 700 distinct values. If you want them treated as categories or as numbers, they need an explicit decision and a cast.