Re: Help with optimizing GROUP BY query

[KimSeong Loh]

Then a full table scan may just be the most efficient way to do it, just make sure you have a big enough sort_buffer.

A full table scan that reads the disk sequentially can be much faster than random access read the rows in sorted order.

Alternative is to use a covering index since you only needs the 2 column userid and score, assuming there are a lot of other columns which will not be useful in this query.