How to optimize query when using sub-queries in left join - mysql

Question

Tables:
Please take a look here to see tables. How to query counting specific wins of team and find the winner of the series
Questions:
How to make query more optimize?
How to reduce query redundancy?
How to make this query more faster?
Summary
As you can see in the example query this part is use many times.
WHERE leagueid = 2096
AND start_time >= 1415938900
AND ((matches.radiant_team_id= 1848158 AND matches.dire_team_id= 15)
OR (matches.radiant_team_id= 15 AND matches.dire_team_id= 1848158))
SELECT matches.radiant_team_id,
matches.dire_team_id,
matches.radiant_name,
matches.dire_name,
TA.Count AS teamA,
TB.Count AS teamB,
TA.Count + TB.Count AS total_matches,
SUM(TA.wins),
SUM(TB.wins),
(CASE
WHEN series_type = 0 THEN 1
WHEN series_type = 1 THEN 2
WHEN series_type = 2 THEN 3
END) AS wins_goal
FROM matches
LEFT JOIN
(SELECT radiant_team_id,
COUNT(id) AS COUNT,
CASE
WHEN matches.radiant_team_id = radiant_team_id && radiant_win = 1 THEN 1
END AS wins
FROM matches
WHERE leagueid = 2096
AND start_time >= 1415938900
AND ((matches.radiant_team_id= 1848158
AND matches.dire_team_id= 15)
OR (matches.radiant_team_id= 15
AND matches.dire_team_id= 1848158))
GROUP BY radiant_team_id) AS TA ON TA.radiant_team_id = matches.radiant_team_id
LEFT JOIN
(SELECT dire_team_id,
COUNT(id) AS COUNT,
CASE
WHEN matches.dire_team_id = dire_team_id && radiant_win = 0 THEN 1
END AS wins
FROM matches
WHERE leagueid = 2096
AND start_time >= 1415938900
AND ((matches.radiant_team_id= 1848158
AND matches.dire_team_id= 15)
OR (matches.radiant_team_id= 15
AND matches.dire_team_id= 1848158))
GROUP BY dire_team_id) AS TB ON TB.dire_team_id = matches.dire_team_id
WHERE leagueid = 2096
AND start_time >= 1415938900
AND ((matches.radiant_team_id= 1848158
AND matches.dire_team_id= 15)
OR (matches.radiant_team_id= 15
AND matches.dire_team_id= 1848158))
GROUP BY series_id
Scheduled Matches
ID| leagueid| team_a_id| team_b_id| starttime
1| 2096| 1848158| 15| 1415938900

Answer 1

I believe it can be done without subqueries.
I made the following match table
And used the following query to group results, one line per series
SELECT
matches.leagueid,
matches.series_id,
matches.series_type,
COUNT(id) as matches,
IF(radiant_team_id=1848158,radiant_name, dire_name) AS teamA,
IF(radiant_team_id=1848158,dire_name, radiant_name) AS teamB,
SUM(CASE
WHEN radiant_team_id=1848158 AND radiant_win=1 THEN 1
WHEN dire_team_id=1848158 AND radiant_win=0 THEN 1
ELSE 0 END) AS teamAwin,
SUM(CASE
WHEN radiant_team_id=15 AND radiant_win=1 THEN 1
WHEN dire_team_id=15 AND radiant_win=0 THEN 1
ELSE 0 END) AS teamBwin
FROM `matches`
WHERE leagueid = 2096
AND start_time >= 1415938900
AND dire_team_id IN (15, 1848158)
AND radiant_team_id IN (15, 1848158)
group by leagueid,series_id,series_type,teamA,teamB
which yields the following result
Please note that, when grouping the results of one series, there isn't such thing as radiant team or dire team. The radiant and dire roles might be switched several times during the same series, so I only addressed the teams as teamA and teamB.
Now, looking at your prior question, I see that you need to determine the series winner based on the series type and each team victories. This would need to wrap the former query and use it as a subquery such as
SELECT matchresults.*,
CASE series_type
WHEN 0 then IF(teamAwin>=1, teamA,teamB)
WHEN 1 then IF(teamAwin>=2, teamA,teamB)
ELSE IF(teamAwin>=3, teamA,teamB)
END as winner
from ( THE_MAIN_QUERY) as matchresults

Answer 2

There may be more efficient ways to get the results you want. But, to make this query more efficient, you can add indexes. This is the repeated where clause:
WHERE leagueid = 2096 AND
start_time >= 1415938900 AND
((matches.radiant_team_id= 1848158 AND matches.dire_team_id= 15) OR
(matches.radiant_team_id= 15 AND matches.dire_team_id= 1848158))
Conditions with or are hard for the optimizer. The following index will be helpful: matches(leagueid, start_time). A covering index (for the where conditions at least) is matches(leagueid, start_time, radiant_team_id, dire_team_id). I would start with this latter index and see if that improves performance sufficiently for your purposes.