**Size:** the dataset has 5 tables and **718,470 rows** in total. The rows break down as: - `raw_kaggle`: 237,205 - `return_kaggle`: 237,196 - `serve_kaggle`: 237,185 - `players_tournament_man_`: 6,422 - `players_man_`: 462 The three match-level tables are the same kind of record: one row per player per match, with different statistics in each. **Columns:** the catalog gives two different counts. - The overview query (`analytics.table_overview`) reports 17, 17, 17, 4 and 3 columns for the five tables above, so **58** in total. - The dataset card lists 21, 21, 21, 8 and 7 columns, so **78** in total. I did not resolve this. I'd treat 58 as the count the overview query returned and 78 as the card's count, and run `inspect_columns` on each table if you need the exact figure. **Time period:** yes, the data is time-stamped. - `Date` is stored as text in all three match tables. After parsing, all 237,205 / 237,196 / 237,185 rows converted cleanly. - Each of the three match tables runs from **13 Aug 2001 to 28 Oct 2024**, about 23 years. - `players_tournament_man_.year` runs from **2002 to 2024**. - `players_man_.playing_years` is a text list of years, which I did not parse. It does not bear on the main range. The card's narrower ranges (for example 1-Apr-2013 to 9-Sep-2024) came from sorting the dates as text. The parsed dates above are the reliable ones.