Question

The goal of this SQL query is to return the user with the most fight wins AND a count of the wins.

I have two tables: users and fights.

User has many fights
Fight belongs to user

Fight has the following columns of concern:

• challenger_id (user_id fk)
• challengee_id (user_id fk)
• challenger_won (boolean)

As you can see, a user can be a challenger or a challengee, but not both.

• If the user is a challenger and the challenger_won = true, then it is considered a win.
• If the user is a challengee and the challenger_won = false, then it is considered a win.
• If the challenger_won = null, then just disregard it.

How would I write the SQL for this?

Answer 1

What about this?

SELECT a.fighter, COUNT(*) AS wins
  FROM (SELECT challenger_id AS fighter
          FROM fights
         WHERE challenger_won = TRUE
        UNION ALL
        SELECT challengee_id AS fighter
          FROM fights
         WHERE challenger_won = FALSE
       ) AS a
 GROUP BY a.fighter;

If challenger_won is NULL, both queries in the UNION will ignore the row, as required.

If you don’t like UNION ALL, you can use, instead:

SELECT a.fighter, SUM(a.wins) AS wins
  FROM (SELECT challenger_id AS fighter, 'R' AS role, COUNT(*) AS wins
          FROM fights
         WHERE challenger_won = TRUE
         GROUP BY fighter, role
        UNION
        SELECT challengee_id AS fighter, 'E' AS role, COUNT(*) AS wins
          FROM fights
         WHERE challenger_won = FALSE
         GROUP BY fighter, role
       ) AS a
 GROUP BY a.fighter;

The first part of the UNION generates rows with ‘R’ as role (for ‘challengeR won’) and the fighter ID and a count of the fights won. The second part generates rows with ‘E’ as role (for ‘challengeE won’) and the fighter ID and a count of the fights won. The role is necessary in case some fighter won 3 fights as challenger and 3 fights as challengee; without the role, the two entries with the same fighter and count would be collapsed into one. You then sum the wins for each fighter in each role.