SQL - Avanceret fjernelse af dubletter ved hjælp af ID NOT IN

Du kan bruge ROW_NUMBER analytisk funktion og korreler sletningen ved hjælp af ROWID pseudo-kolonne:

DELETE FROM my_table
WHERE ROWID IN (
  SELECT ROWID
  FROM   (
    SELECT ROW_NUMBER() OVER (
             PARTITION BY student_id, subject_id, class_id
             ORDER BY expertise_lvl DESC
           ) AS rn
    FROM   my_table
  )
  WHERE  rn > 1
)

db<>fiddle her