count and group by first n(minimum) items in group - mysql

Question

I've been through several of the "n from M" type solutions and not been able to get close to what I'm after though it's possible that the question has been asked before in some other format.
I've tried examples from this MySQL Group By with top N number of each kind and this http://www.xaprb.com/blog/2006/12/07/how-to-select-the-firstleastmax-row-per-group-in-sql/ none of which appear to apply to what I'm trying to do.
What I'm trying to do is determine the best teams in a running race, individual runners aren't a problem, gender, age categories can be taken care of. The rules for team prizes are based on membership of a club.
Clubs must have at least 3 runners to qualify for the team competition.
Only the first 3 runners from each club count towards the competition.
The team position is determined by the sum of the qualifying runners so runners from club A who finish 2nd 9th & 10th get 21pts, runners from club B who finish 4th, 5th & 6th get 15pts, etc.
I've a table with the following fields:
+---------------+-------------+------+-----+---------+----------------+
| Field | Type | Null | Key | Default | Extra |
+---------------+-------------+------+-----+---------+----------------+
| id | int(11) | NO | PRI | NULL | auto_increment |
| runner_id | int(11) | YES | | NULL | |
| club_id | int(11) | YES | | NULL | |
| race_id | int(11) | YES | | NULL | |
| race_number | int(11) | YES | | NULL | |
| category | varchar(20) | YES | | NULL | |
| finish_time | int(11) | YES | | NULL | |
| race_position | int(11) | YES | | NULL | |
+---------------+-------------+------+-----+---------+----------------+
Only club_id and race_position are relevant to the query. runner_id, club_id and race_id are foreign keys and I need to be able to extract data (given_name, family_name, age, club_name, etc.) from those tables when creating the results.
This is typical data:
+----+-----------+---------+---------+-------------+-----------+-------------+---------------+
| id | runner_id | club_id | race_id | race_number | category | finish_time | race_position |
+----+-----------+---------+---------+-------------+-----------+-------------+---------------+
| 53 | 26 | 1 | 85 | 17 | Msenior | 1666 | 11 |
| 35 | 39 | 1 | 85 | 4 | Munder_18 | 1503 | 4 |
| 63 | 61 | 2 | 85 | 27 | Mvet_50 | 1610 | 9 |
| 42 | 46 | 2 | 85 | 11 | Lvet_40 | 1773 | 14 |
| 38 | 42 | 2 | 85 | 7 | Lunder_18 | 1793 | 17 |
| 56 | 36 | 9 | 85 | 20 | Msenior | 1561 | 6 |
| 44 | 48 | 9 | 85 | 13 | Msenior | 1667 | 12 |
| 64 | 62 | 9 | 85 | 28 | Msenior | 1660 | 10 |
| 49 | 52 | 9 | 85 | 18 | Msenior | 1432 | 1 |
| 47 | 51 | 10 | 85 | 16 | Msenior | 1779 | 15 |
| 61 | 59 | 11 | 85 | 25 | Mvet_50 | 1502 | 3 |
| 33 | 38 | 11 | 85 | 2 | Munder_18 | 1440 | 2 |
| 65 | 63 | 11 | 85 | 29 | Mvet_40 | 1566 | 8 |
| 54 | 54 | 12 | 85 | 19 | Msenior | 1785 | 16 |
| 58 | 56 | 12 | 85 | 23 | Msenior | 1546 | 5 |
| 37 | 41 | 12 | 85 | 6 | Munder_18 | 1668 | 13 |
| 45 | 49 | 14 | 85 | 14 | Mvet_50 | 1565 | 7 |
+----+-----------+---------+---------+-------------+-----------+-------------+---------------+
What I want to end up with is this:
+----+-----------+---------+---------+-------------+-----------+-------------+---------------+
| id | runner_id | club_id | race_id | race_number | category | finish_time | race_position |
+----+-----------+---------+---------+-------------+-----------+-------------+---------------+
| 33 | 38 | 11 | 85 | 2 | Munder_18 | 1440 | 2 |
| 61 | 59 | 11 | 85 | 25 | Mvet_50 | 1502 | 3 |
| 65 | 63 | 11 | 85 | 29 | Mvet_40 | 1566 | 8 |
| 49 | 52 | 9 | 85 | 18 | Msenior | 1432 | 1 |
| 56 | 36 | 9 | 85 | 20 | Msenior | 1561 | 6 |
| 64 | 62 | 9 | 85 | 28 | Msenior | 1660 | 10 |
| 58 | 56 | 12 | 85 | 23 | Msenior | 1546 | 5 |
| 37 | 41 | 12 | 85 | 6 | Munder_18 | 1668 | 13 |
| 54 | 54 | 12 | 85 | 19 | Msenior | 1785 | 16 |
| 63 | 61 | 2 | 85 | 27 | Mvet_50 | 1610 | 9 |
| 42 | 46 | 2 | 85 | 11 | Lvet_40 | 1773 | 14 |
| 38 | 42 | 2 | 85 | 7 | Lunder_18 | 1793 | 17 |
+----+-----------+---------+---------+-------------+-----------+-------------+---------------+
So even though runner_id of 52 won the race, he wasn't in the winning team.
I'm running all this under Codeigniter/Datamapper ORM but I can pass a full SQL query string down through this layer.
I hope all this makes sense.

Answer 1

MySQL lacks important features to get this solved (CTEs, window functions) but you can workaround them with some user defined variables and by paying the performance cost:
SELECT s1.id, s1.runner_id, s1.club_id, s1.race_id, s1.race_number, s1.category,
s1.finish_time, s1.race_position
FROM (
SELECT t1.*,
#club_rank := if(#prev_club = t1.club_id, #club_rank + 1, 1) club_rank,
#prev_club := t1.club_id
FROM t t1
CROSS JOIN (SELECT #prev_club := NULL, #club_rank := 1) init
ORDER BY t1.club_id, t1.race_position
) s1
JOIN (
SELECT club_id, count(*) teamSize, sum(race_position) teamPosition FROM t
GROUP BY club_id
) s2 ON s1.club_id = s2.club_id
WHERE club_rank <= 3 AND teamSize >= 3
ORDER BY teamPosition, race_position
Output:
| ID | RUNNER_ID | CLUB_ID | RACE_ID | RACE_NUMBER | CATEGORY | FINISH_TIME | RACE_POSITION |
|----|-----------|---------|---------|-------------|-----------|-------------|---------------|
| 33 | 38 | 11 | 85 | 2 | Munder_18 | 1440 | 2 |
| 61 | 59 | 11 | 85 | 25 | Mvet_50 | 1502 | 3 |
| 65 | 63 | 11 | 85 | 29 | Mvet_40 | 1566 | 8 |
| 49 | 52 | 9 | 85 | 18 | Msenior | 1432 | 1 |
| 56 | 36 | 9 | 85 | 20 | Msenior | 1561 | 6 |
| 64 | 62 | 9 | 85 | 28 | Msenior | 1660 | 10 |
| 58 | 56 | 12 | 85 | 23 | Msenior | 1546 | 5 |
| 37 | 41 | 12 | 85 | 6 | Munder_18 | 1668 | 13 |
| 54 | 54 | 12 | 85 | 19 | Msenior | 1785 | 16 |
| 63 | 61 | 2 | 85 | 27 | Mvet_50 | 1610 | 9 |
| 42 | 46 | 2 | 85 | 11 | Lvet_40 | 1773 | 14 |
| 38 | 42 | 2 | 85 | 7 | Lunder_18 | 1793 | 17 |
Fiddle here.

Answer 2

A bit late as I've been indisposed.
I came up with an inelegant solution. I added a club_total column to the table. I then loop over the table with one query for each club getting the first N runners with a query like:
select * from entries where race_id=? and club_id=? LIMIT ? order by race_position;
I then ignore those clubs with less than N finishers and sum the race positions of the other clubs and write this value back to the table.
Finally I run another query to extract only those rows with club totals:
select * from entries where club_total > 0 and race_id=? order by club_total, race_position;
Like I said, it's not elegant and it's certainly not quick (I've not timed it) but it's only going to get run a handful of times a year on one machine and the record set is a couple of hundred rows at a maximum. With a small data set it's not appreciably slower than a simple query with the data being displayed via AJAX. Getting the job done in this case is more important than speed. I wouldn't use this method for any situation where performance was an issue