Mysql Calculate rank of teams from different rows - mysql

Question

I'm trying to build a kind of peddy paper. For that I have the following tables:
teams
CREATE TABLE `teams` (
`id` int(11) NOT NULL AUTO_INCREMENT,
`creator_id` int(11) NOT NULL,
`friend_id` int(11) DEFAULT NULL,
`team_name` varchar(128) NOT NULL,
PRIMARY KEY (`id`)
);
team_log
CREATE TABLE IF NOT EXISTS `progress_tracker` (
`id` int(8) NOT NULL AUTO_INCREMENT,
`user_id` int(8) NOT NULL,
`team_id` int(11) NOT NULL,
`date` date NOT NULL,
`clues_found` int(11) NOT NULL,
`clues_to_find` int(11) NOT NULL,
PRIMARY KEY (`id`)
);
Each team is composed by two users;
Each user starts out with a variable number of clues found;
clues_found can either increase or decrease. No guarantee that the highest number is the latest;
I need to get a rank of the teams (in percentage) based on the average of the number of clues the user found since they joined (for both users in a team) - clues_found on the row with biggest date minus clues_found on the record with lowest date).
For instance if I have the following data for each table:
teams table data
+--------+------------+------------+---------------+
| id | creator_id | friend_id | team_name |
+--------+------------+------------+---------------+
| 1 | 25 | 28 | Test1 |
| 2 | 31 | 5 | Test2 |
+--------+------------+------------+---------------+
team_log table data
+--------+---------+---------+------------+-------------+---------------+
| id | user_id | team_id | date | clues_found | clues_to_find |
+--------+---------+---------+------------+-------------+---------------+
| 1 | 25 | 1 | 2013-01-6 | 3 | 24 |
| 2 | 25 | 1 | 2013-01-8 | 7 | 24 |
| 3 | 25 | 1 | 2013-01-10 | 10 | 24 |
| 4 | 28 | 1 | 2013-01-8 | 5 | 30 |
| 5 | 28 | 1 | 2013-01-14 | 20 | 30 |
| 6 | 31 | 2 | 2013-01-11 | 6 | 14 |
| 7 | 5 | 2 | 2013-01-9 | 2 | 20 |
| 8 | 5 | 2 | 2013-01-10 | 10 | 20 |
| 9 | 5 | 2 | 2013-01-12 | 14 | 20 |
+--------+---------+---------+------------+-------------+---------------+
Desired Result
+-------------+---------------------+
| team_id | team_percentage |
+-------------+---------------------+
| 1 | 39,58333333 |
| 2 | 30 |
+-------------+---------------------+
As a reference this is an intermediate representation which might help to understand:
+-------------+---------+---------------------+
| user_id | team_id | precentage_per_user |
+-------------+---------+---------------------+
| 25 | 1 | 29,16666667 |
| 28 | 1 | 50 |
| 31 | 2 | 0 |
| 5 | 2 | 60 |
+-------------+---------+---------------------+
So far I have the following sql:
SELECT STRAIGHT_JOIN
tl2.team_id, (tl2.weight - tl1.weight)*100/tl2.clues_to_find
from
( select
team_id,user_id,clues_found
FROM
`team_log`
where 1
group by
team_id, user_id
order by
`date` ) base
join (select team_id, user_id, clues_found, clues_to_find from `team_log` where user_id = base.user_id and team_id = base.team_id group by team_id, user_id order by `date` desc) tl2
But this returns an error as I'm not allowed to use base.user_id inside the second query. I'm also not very sure I'm heading in the right direction.
Can anyone help please?

Answer 1

Here's another query that will produce the correct result:
SELECT calc.team_id, AVG((calc.end_clues - calc.start_clues)/calc.total_clues*100) as team_percentage
FROM
(SELECT log1.user_id, log1.team_id, log1.clues_found as start_clues, log2.clues_found as end_clues, log2.clues_to_find as total_clues FROM team_log log1
JOIN
(SELECT MIN(id) as start_id, MAX(id) as end_id FROM team_log GROUP BY user_id) ids
ON ids.start_id = log1.id
JOIN team_log log2 ON ids.end_id = log2.id) calc
GROUP BY team_id
ORDER BY team_id;
And the SQL Fiddle-link...

Answer 2

Please take a look at this and comment:
SQLFIDDLE DEMO
Team pct:
select z.team_id, avg(z.pct) as teampct
from (
select x.user_id, y.team_id, x.mndate,
y.mxdate, x.mnclues_found,
y.mxclues_found,
(((y.mxclues_found - x.mnclues_found)*100)
/y.mxclues_tofind) pct
from
(select user_id, team_id, min(date) mndate,
min(clues_found) as mnclues_found
from team_log
group by user_id, team_id) x
left join
(select user_id, team_id, max(date) mxdate,
max(clues_found) as mxclues_found,
max(clues_to_find) as mxclues_tofind
from team_log
group by user_id, team_id) y
on x.user_id = y.user_id and
x.team_id = y.team_id) z
group by z.team_id
;
Results 1:
| USER_ID | TEAM_ID | MNDATE | MXDATE | MNCLUES_FOUND | MXCLUES_FOUND | PCT |
-------------------------------------------------------------------------------------
| 5 | 2 | 13-01-09 | 13-01-12 | 2 | 14 | 60 |
| 25 | 1 | 13-01-06 | 13-01-10 | 3 | 10 | 29.1667 |
| 28 | 1 | 13-01-08 | 13-01-14 | 5 | 20 | 50 |
| 31 | 2 | 13-01-11 | 13-01-11 | 6 | 6 | 0 |
Results final:
| TEAM_ID | TEAMPCT |
----------------------
| 1 | 39.58335 |
| 2 | 30 |

Answer 3

This is a bit ugly, but should work:
select
team_id,
AVG(percentage_per_user) as team_percentage
from (select
team_id,
user_id,
((select clues_found from progress_tracker as x
where x.user_id = m.user_id order by x.date desc limit 0, 1)
- (select clues_found from progress_tracker as y
where y.user_id = m.user_id order by y.date asc limit 0, 1))
/ MAX(clues_to_find)
as percentage_per_user
from progress_tracker as m
group by team_id, user_id
) as userScore
group by team_id
order by team_percentage desc;
Note the inner query run by itself will yield your intermediate "per-user" result.

Answer 4

SQLFiddle
SELECT `team_id`,
(SUM(CASE WHEN b.`date` IS NULL THEN 0 ELSE `clues_found` * 100 / `clues_to_find` END) -
SUM(CASE WHEN c.`date` IS NULL THEN 0 ELSE `clues_found` * 100 / `clues_to_find` END)) / 2
FROM `team_log` a
LEFT JOIN (
SELECT `team_id`, `user_id`, MAX(date) AS `date`
FROM `team_log`
GROUP BY `team_id`, `user_id`) b
USING (`team_id`, `user_id`, `date`)
LEFT JOIN (
SELECT `team_id`, `user_id`, MIN(date) AS `date`
FROM `team_log`
GROUP BY `team_id`, `user_id`) c
USING (`team_id`, `user_id`, `date`)
GROUP BY `team_id`
Since you say there are always two team members, I've used /2. It would be slightly more complex for variable-sized teams.