The Short Answer

To build a football scouting database, store the raw event rows first, then build simple tables around them: players, teams, matches and events with pitch coordinates and qualifiers. Keep every event as its own row, add player and team tables you can join to it, then write queries that turn those rows into shortlists. That structure answers new scouting questions without rebuilding your notes each time.

What does a football scouting database actually need?

A scouting database has one job: answer a new question about a player without starting the research again. Spreadsheets fail at that job because each note lives in its own file, written for one question, in one scout's words. A database fixes it by storing the underlying facts once, in a shape any question can reuse.

The facts that matter are events. In Dribble's record of Crystal Palace against Arsenal on 24 May 2026 there are 1,574 events in a single match, and 920 of them are passes. Shots account for just 25. A scout who only stores the 25 shots throws away the build up, the pressing and the positioning that explain why those shots happened. Store the 1,574 rows and the match can be recut for any question: a winger's touches in the final third, a midfielder's passes under pressure, a full back's duels on one side of the pitch.

Scale is the second reason. Dribble's BigQuery dataset holds more than 57 million event rows across 32 competitions from 2020 onwards, alongside 90,582 players and 41,017 matches. No spreadsheet survives contact with that volume. A database is not a bigger spreadsheet; it is the tool that makes event data usable at all.

Which tables should you build first?

Start with four tables. Everything else is a view or a join on top of them.

Table What it holds Why it matters
players One row per player: name, position, date of birth, nationality, current team Age and position filters live here, so a query can ask for under 23 midfielders instead of every midfielder
teams One row per team, with competition and season Lets you compare a player against his league, not against world football
matches One row per match: date, home and away team, score, competition The calendar and the context: form, strength of schedule, home and away splits
match_events One row per event: match, player, team, type, outcome, pitch coordinates, qualifiers The evidence itself, from which every scouting metric is counted

Keep the raw event rows untouched. In Dribble's match_events, richer tiers carry more of each event: Starter has 19 columns, Pro has 25 and Max has 31. The six Max only columns are period_id, goalmouth_y, goalmouth_z, blocked_x, blocked_y and qualifiers_json. Build your scouting views on top of the raw table, so that when a new question arrives you write a new view instead of editing the evidence.

The subscriber workspace shows the full schema with sample files for player_matches and match_events, which is the fastest way to see the shape before you commit to a design.

How do you model event data for scouting questions?

Model around the questions scouts actually ask. Most of them have the same shape: which players, in which competitions, over which period, did a specific thing more often or better than their peers?

Three rules make those questions easy to write. First, one row per event, always. The moment you store a match as totals, you have decided in advance which questions can be asked. Second, keep coordinates raw. An event carries x and y for where it happened, and passes carry end_x and end_y for where they arrived. Zones, thirds and boxes are then definitions you apply in a query, and you can change the definition without recollecting the data. Third, keep qualifiers attached to the event they describe. A qualifier set can record that a chance was assisted, that a pass was a through ball, or that a shot was a big chance, in the event's own qualifiers_json field. That is how a shortlist query can separate a striker who creates big chances from one who merely shoots often.

Our guide to football data in BigQuery walks through how those tables sit inside BigQuery, where the scouting database can live in your own Google Cloud project and be queried with plain SQL.

What does a shortlist query look like in practice?

Here is a starting query on match_events, using the columns exactly as they appear in the sample data. It finds players whose passing repeatedly reaches the final fifth of the pitch, a simple first pass at creative output:

SELECT
  player,
  team,
  COUNT(DISTINCT match_id) AS matches_seen,
  COUNTIF(type = 'PASS' AND end_x >= 80) AS passes_into_final_fifth,
  COUNTIF(type = 'PASS' AND is_through_ball) AS through_balls,
  COUNTIF(is_assist) AS assists
FROM match_events
WHERE player IS NOT NULL
GROUP BY player, team
HAVING matches_seen >= 5
ORDER BY passes_into_final_fifth DESC
LIMIT 25

Read it as a scout would. The pitch runs from 0 to 100, so a pass ending at end_x of 80 or more finished in the final fifth. The matches_seen floor keeps one good game from topping the list. The result is not a verdict; it is a watchlist of 25 names, each one traceable back to the individual event rows behind the count, ready for video checks and live scouting. Change the filters and the same table answers the next question: add an age filter from the players table for an under 23 list, or restrict the competitions to the two leagues your club actually recruits from.

That traceability is the difference between a database and a ranking someone sent you. Every number on the shortlist can be opened up into the passes, matches and minutes that produced it.

How do you keep the database useful after week one?

Three habits keep a scouting database alive. Refresh on a schedule, not on request: Dribble's BigQuery tiers refresh every 2 weeks on Starter at $49 a month, every 72 hours on Pro at $149 a month, and daily on Max at $499 a month, so pick the rhythm your recruitment calendar needs and let the tables update themselves. Keep a saved library of the queries your scouts reuse, with the question written above each one in plain words, so a new analyst inherits the method and not just the output. And review the shortlist against outcomes twice a season: which flagged players kicked on, which were false positives, and which filter would have separated them?

Start small. Load one competition, build the four tables, write the shortlist query above, and check its output against players your scouts already rate. When the database agrees with the eye on the known names and surfaces two or three unknown ones worth watching, it has earned the right to grow.

Frequently asked questions

What is a football scouting database? A structured store of players, teams, matches and event rows that lets a scout answer new questions with queries instead of starting research again. The event rows are the core: every pass, shot and tackle kept as its own row with pitch coordinates, so any metric or shortlist can be rebuilt from the evidence.

What data do I need for player scouting analysis? At minimum: a players table with position and date of birth, a matches table for context, and event data with XY coordinates and qualifiers. Dribble covers 32 competitions from 2020 onwards, with more than 57 million event rows, 90,582 players and 41,017 matches in BigQuery.

Do I need event data, or will match statistics do? Match statistics are totals counted from events by someone else, so they only answer the questions the provider chose in advance. Event data keeps the underlying rows: in one Crystal Palace against Arsenal match there are 1,574 events and only 25 shots. If your questions change each window, you need the rows, not just the totals.

How much does it cost to build a scouting database with Dribble? BigQuery access starts at $49 a month on Starter, $149 a month on Pro and $499 a month on Max, with refreshes every 2 weeks, every 72 hours and daily respectively. The match_events table carries 19 columns on Starter, 25 on Pro and 31 on Max. Free sample data is available on Kaggle so you can test the schema before paying.

The event counts in this article come from Dribble's Max tier sample of Crystal Palace v Arsenal, 24 May 2026 (1,574 events). The dataset totals are Dribble's published BigQuery figures: more than 57 million event rows, 90,582 players and 41,017 matches across 32 competitions.

Build your scouting database with Dribble data