The sample above is 10 rows from `raw.raw_kaggle`, the main match-level fact table (237,205 rows, one row per player per match). I haven't sampled the other four tables. **What the rows show** - **Coverage:** the rows span 2009 to 2024. They cover Masters events (Miami, Monte Carlo, Paris), Grand Slams (Wimbledon, Roland Garros), and lower-tier Challenger and M15 events (Cuernavaca CH, M15 Loule, Slovakia CH). - **Surfaces and rounds:** the sample includes Hard, Clay, and Grass surfaces. Rounds range from R128 and R64 through R16, QF, and F, plus a qualifying round (Q2). - **Rankings:** `Rk` and `vRk` are the player's ranking and the opponent's ranking. They run from top-10 players (Monfils at 10 vs. Federer at 2) to players ranked in the 600s (Etcheverry at 645 vs. 580). - **Match stats:** `TP` (total points), `Aces`, `DFs` (double faults), `SP` (serve points), `1SP` and `2SP` (first- and second-serve points) and `vA` are filled for most rows. Dimitrov's Miami final, for example, was a 6-3 6-1 match of 93 points with 6 aces and 1 double fault. **Data quality** - **Missing stats:** the Etcheverry M15 Loule row has blank `TP`, `Aces`, `DFs`, `SP`, `1SP`, `2SP`, `vA` and `Time`. Lower-tier matches seem to lack detailed stats. `Time` is also blank on the Kovalik row. - **Garbled `against` column:** it holds a messy list of both players' names, and the names are sometimes split mid-word. Examples are `['San', 'oEhrat[SUI]', 'Etcheverry']` and `['(2)RogerFe', 'rer[SUI]', '(11)Monfils']`. The opponent name can't be used directly from this column without cleaning. - **Seed and country tags:** entries carry seed and country markers such as `(2)`, `[ITA]` and `(WC)`. - **Ingestion metadata:** the trailing `_ingestion_timestamp`, `_batch_id`, `_source_file` and `_source_system` columns are pipeline metadata, not tennis data. The other four tables are `raw.serve_kaggle`, `raw.return_kaggle`, `raw.players_tournament_man_` and `raw.players_man_`. I can sample any of them next.