Postgres OFFSET the results of subqueries group by - sql

Question

Im calculating weekly score for game where certain weeks there are bonuses, and when totaling your score the lowest two scores are dropped.
id
name
week
score
1
Player A
1
10
2
Player A
2
20
3
Player A
3
30
4
Player A
4
40
5
Player B
1
5
6
Player B
2
10
7
Player B
3
15
8
Player B
4
20
Let's say in week 2 your score should be doubled,
So A's scores should be [10,40,30,40] and B [5,20,15,20]
With the rules of removing the two lowest scores
A [40,40] total 80
B [20,20] total 40
If I run this this query
select name, sum(special_scores) as total_score
from(
select
name,
case
when week = 2 then score * 2
else score
end special_scores
from public.standings
where name = 'Player A'
order by special_scores
offset 2
) s
group by name
order by total_score desc;
I see the expected result of totaling the score column and omitting the last two results, so I believe my sub query is correct.
However if I remove the where clause from the subquery
select name, sum(special_scores) as total_score from (
select name, case
when week = 2 then score * 2
else score
end special_scores from public.standings
order by special_scores
offset 2
) s
group by name
order by total_score desc
The table will populate but will not omit the two lowest scores
So I'm getting something like
name.
total_score
Player A
120
Player B
60
Could someone help as to why the offset in the second query is not removing the scores before totaling?

Answer 1

Brake the problem into subproblems.
First create a subquery that computes the actual scores (when they have to be doubled). keep id, week and actual score.
THen use a window function (row_number()) to drop the bottom two scores.
Then sum the results by id.
Finally, join the id to this result table to get the player name too.