Joining two tables twice - mysql

Question

Writing a query for a basketball database, our table for games has as a winnerID and a loserID, each being a teamID. Tried the following two queries, each correctly giving me the number of wins but giving me the same number for losses.
SELECT team.name as Team_Name, COUNT(team.teamID=winner.winnerID) as Wins, COUNT(team.teamID=loser.loserID) as Losses
FROM team join games winner on winner.winnerID=team.teamID join games loser on loser.loserID=team.teamID
GROUP BY team.name
ORDER BY Wins, Team_Name;
SELECT team.name as Team_Name, COUNT(team.teamID=games.winnerID) as Wins, COUNT(team.teamID=games.loserID) as Losses
FROM (team INNER JOIN games on games.winnerID=team.teamID)
GROUP BY team.name
ORDER BY Wins, team.name;
Help?
EDIT: Forgot to mention, purpose of query is to get number of wins and number of losses of each team.

Answer 1

The COUNT aggregate gets a count of non-NULL values. That means it includes ones and zeros.
Evaluated in a numeric context, an equality comparison returns 1 for TRUE and returns 0 for FALSE, and only returns NULL if either side (or both sides) is NULL.
To add up the ones and ignore the zeros, you could use a SUM aggregate instead.
One of the big problems with the query is the potential to return duplicates, due to the cross join between winner and loser. If a team has 5 wins and 4 losses, the query is going to generate an intermediate set of 20 (= 5 x 4) rows.
To get a count of wins and losses from that, we'd need a unique identifier for a game (for example, a gameid column in the game table that is the PRIMARY KEY.) With that , we could get a count of distinct values of gameid. For example:
COUNT(DISTINCT IF(winner.winnerid=team.teamid,winner.gameid,NULL)) AS wins
There are several query patterns that will get you the number of wins and the number of losses for each team.
Here's one example:
SELECT t.name AS team_name
, COUNT(IF(t.teamid=g.winnerid,1,NULL) AS wins
, COUNT(IF(t.teamid=g.loserid ,1,NULL) AS losses
FROM team
LEFT
JOIN games g
ON ( g.winnerid = t.teamid OR g.loserid = t.teamid )
GROUP
BY t.name
ORDER
BY wins DESC
, t.name
With this, we are only joining to the games table once, so we won't get a cross product. Also note that if teamid is not equal to winnerid, we return a NULL instead of a 0. So a COUNT will include only winners, not all of the rows.
We use an outer join (rather than an inner join), in case there are no related rows in games for the team. That allows us to return a team with counts of zero.
We could use a SUM aggregate instead of a COUNT, For example:
SELECT t.name AS team_name
, IFNULL(SUM(t.teamid=g.winnerid),0) AS wins
, IFNULL(SUM(t.teamid=g.loserid) ,0) AS losses
FROM team
LEFT
JOIN games g
ON ( g.winnerid = t.teamid OR g.loserid = t.teamid )
GROUP
BY t.name
ORDER
BY wins DESC
, t.name
With a SUM() we have a potential to return NULL values. To get those converted to zeros, we use an IFNULL function.
And there are several other query patterns that would get you an equivalent result... e.g. instead of using joins, use correlated subqueries in the SELECT list...
SELECT t.name
, ( SELECT COUNT(1)
FROM game w
WHERE w.winner_id = t.teamid
) AS wins
, ( SELECT COUNT(1)
FROM game l
WHERE l.loserid = t.teamid
) AS losses
FROM team t
ORDER BY wins DESC
, t.name