Question

I am a fairly experienced SQL Server developer but this problem has me REALLY stumped.

I have a FUNCTION. The function is referencing a table that is something like this…

PERFORMANCE_ID, JUDGE_ID, JUDGING_CRITERIA, SCORE  
--------------------------------------------------  
101, 1, 'JUMP_HEIGHT', 8   
101, 1, 'DEXTERITY', 7  
101, 1, 'SYNCHRONIZATION', 6  
101, 1, 'SPEED', 9  
101, 2, 'JUMP_HEIGHT', 6   
101, 2, 'DEXTERITY', 5  
101, 2, 'SYNCHRONIZATION', 8  
101, 2, 'SPEED', 9  
101, 3, 'JUMP_HEIGHT', 9   
101, 3, 'DEXTERITY', 6  
101, 3, 'SYNCHRONIZATION', 7  
101, 3, 'SPEED', 8  
101, 4, 'JUMP_HEIGHT', 7   
101, 4, 'DEXTERITY', 6  
101, 4, 'SYNCHRONIZATION', 5  
101, 4, 'SPEED', 8  

In this example there are 4 judges (with IDs 1, 2, 3, and 4) judging a performance (101) against 4 different criteria (JUMP_HEIGHT, DEXTERITY, SYNCHRONIZATION, SPEED).
(Please keep in mind that in my real data there are 10+ criteria and at least 6 judges.)

I want to aggregate the results in a score BY JUDGING_CRITERIA and then aggregate those into a final score by summing…something like this…

SELECT SUM (Avgs) FROM
(SELECT AVG(SCORE) Avgs 
  FROM PERFORMANCE_SCORES
  WHERE PERFORMANCE_ID=101
  GROUP BY JUDGING_CRITERIA) result 

BUT… that is not quite what I want IN THAT I want to EXCLUDE from the AVG the highest and lowest values for each JUDGING_CRITERIA grouping. That is the part that I can’t figure out. The AVG should be applied only to the MIDDLE values of the GROUPING FOR EACH JUDGING_CRITERIA. The HI value and the LO value for JUMP_HEIGHT should not be included in the average. The HI value and the LO value for DEXTERITY should not be included in the average. ETC.

I know this could be accomplished with a cursor to set the hi and lo for each criteria to NULL. But this is a FUNCTION and should be extremely fast.

I am wondering if there is a way to do this as a SET operation but still automatically exclude HI and LO from the aggregation?

Thanks for your help. I have a feeling it can probably be done with some advanced SQL syntax but I don’t know it.

One last thing. This example is actually a simplification of the problem I am trying to solve. I have other constraints not mentioned here for the sake of simplicity.

Seth

Answer 1

EDIT: -Moved the WHERE clause to inside the CTE.
-Removed JudgeID from the partition

This would be my approach

;WITH Agg1 AS 
(
    SELECT   PERFORMANCE_ID
            ,JUDGE_ID
            ,JUDGING_CRITERIA
            ,SCORE
            ,MinFind    = ROW_NUMBER() OVER (   PARTITION BY     PERFORMANCE_ID
                                                                ,JUDGING_CRITERIA
                                                ORDER BY SCORE ASC  )
            ,MaxFind    = ROW_NUMBER() OVER (   PARTITION BY     PERFORMANCE_ID
                                                                ,JUDGING_CRITERIA
                                                ORDER BY SCORE DESC )
    FROM PERFORMANCE_SCORES
    WHERE PERFORMANCE_ID=101
)
SELECT AVG(Score)
FROM Agg1
WHERE MinFind > 1
AND MaxFind > 1
GROUP BY JUDGING_CRITERIA