In step 1, we slim the dataset down to concentrate on only the columns of importance. This recipe would work the same with the entire DataFrame. Step 2 shows how to sort a DataFrame by a single column, which is not exactly what we wanted. Step 3 sorts multiple columns at the same time. It works by first sorting all of title_year and then, within each distinct value of title_year, sorts by imdb_score.

The default behavior of the drop_duplicates method is to keep the first occurrence of each unique row, which would not drop any rows as each row is unique. However, the subset parameter alters it to only consider the column (or list of columns) given to it. In this example, only one row for each year will be returned. As we sorted by year and score in the last step, the highest scoring movie for each year is what we get.