Question

I have this table

select * from points

+---------+------+------+
| NAME    | Type | RANK |
+---------+------+------+
| A       | H    |  90  |
| A       | M    | 100  |
| A       | H    | N/A  |
| A       | H    | N/A  |
| A       | H    | N/A  |
| B       | H    | 100  |
| B       | M    | 100  |
| B       | L    | 100  |
| C       | H    |  85  |
| C       | M    | 100  |
+---------+------+------+

I’m using this query

SELECT name,
       CAST(
       (      -- only have H, or only have M, or only have L:
         CASE WHEN  `# of H` = 0  AND  `# of M` = 0  THEN  100 * `# of active L` / `# of L`
              WHEN  `# of H` = 0  AND  `# of L` = 0  THEN  100 * `# of active M` / `# of M`
              WHEN  `# of M` = 0  AND  `# of L` = 0  THEN  100 * `# of active H` / `# of H`
              -- only have H & M, or only have H & L, or only have M & L:
              WHEN  `# of H` = 0  THEN  60 * `# of active M` / `# of M` + 40 * `# of active L` / `# of L`
              WHEN  `# of M` = 0  THEN  90 * `# of active H` / `# of H` + 20 * `# of active L` / `# of L`
              WHEN  `# of L` = 0  THEN  80 * `# of active H` / `# of H` + 20 * `# of active M` / `# of M`
              -- have all three:
              ELSE  70 * `# of active H` / `# of H` + 20 * `# of active M` / `# of M` + 10 * `# of active L` / `# of L`
         END
       ) AS SIGNED ) AS score
  FROM ( SELECT name,
                SUM(IF(         type = 'H', 1, 0))  AS `# of H`,
                SUM(IF(rank AND type = 'H', 1, 0))  AS `# of active H`,
                SUM(IF(         type = 'M', 1, 0))  AS `# of M`,
                SUM(IF(rank AND type = 'M', 1, 0))  AS `# of active M`,
                SUM(IF(         type = 'L', 1, 0))  AS `# of L`,
                SUM(IF(rank AND type = 'L', 1, 0))  AS `# of active L`
           FROM points
          GROUP BY name
       ) t
 ORDER
    BY name
;

I get this Output

+---------+-------+
| NAME    | SCORE |
+---------+-------+
| A       |   60  | <--[(2xH)=40 + (1xM)=20] =60
| B       |  100  | <--[(1xH)=70 + (1xM)=20 + (1xL)=10] =100
| C       |  100  | <--[(1xH)=80 + (1xM)=20] =100
+---------+-------+

I need this Desired output

+---------+-------+
| NAME    | SCORE |
+---------+-------+
| A       |   36  | <--[70/4=(17.5 per H) therefore (17.5)*(rank of that h: 90%)=15.75 + (M values, which equals 20/1 =20 Therefore: rank of that m:100% * 20 = 100) = 36 rounded
| B       |  100  | <--[(1xH)=70 + (1xM)=20 + (1xL)=10] =100
| C       |   88  | <--[(1xH)=80 + (1xM)=20] =100
+---------+-------+

Computations required:

• Type can have only three values: {H, M, L};

• When all values are present, they are graded as followed:

  H=70 M=20 L=10

• If an name has more than one kind of Type (H, M, or L) then points are distributed as followed:

• H/(number of H) ; M/(number of M); L/(number of L)

— Example: A has 4 H therefore 70 / 4 = 17.5 for each H

• But some names have a complete set with out having all ‘Types.
  — example : C has Type values: ‘H&M` only

• Now Type ‘H’ and ‘M’ have to equal 100 for C.

So when only ‘H` and ‘M’ are present they are graded as followed:

H=80 M=20

• Equally if another animal comes along with only two Type values M & L they will be graded as followed:

M=60 L=40

• Equally if another animal comes along with only two Type values H & L they will be graded as followed:

H=90 L=10

And also

• if only H is presnet H=100

• if only M is presnet M=100

• if only L is presnet L=100

Answer 1

This looks familiar. 🙂

Your description is inconsistent in a number of places — for example, your “desired output” for A uses 70 and 20, even though no As have type L — but if you mean what I think you do, then the main change you need is to change SUM(IF(rank AND type = 'H', 1, 0)) AS `# of active H` to SUM(IF(type = 'H', rank / 100.0, 0)) AS `rank of H` (and likewise for M and L), and change all references to `# of active H` to refer to `rank of H` instead. This way each record will be included in proportion to its rank, rather than being an all-or-nothing thing.

You’ll also want to use ROUND instead of CAST — or in addition to CAST — when converting your score to an integer.