If you can identify a "primary key" value in the data that you're putting into the table -- e.g. if the combination of "round" and "playername" should never repeat in the table -- then you could start with a query that pulls out all the existing "round,playername" tuples in the table, and store those as keys to a hash (values assigned to those keys won't matter, just so long as the keys are made to exist).

Then, use the "prepare" method in DBI to create two SQL statement handles -- one for update, and one for insert.

Now, as you go through the "weekly_file", check each "round,playername" value to see if it exists as a hash key from that initial query -- if it's there, use the update statement handle, otherwise, use the insert.

I wrote a wrapper for DBI that makes this sort of thing fairly simple to code in Perl (the pod even gives an example a bit like this) -- I'm not claiming it's the "best" or "optimal" solution, but it works for me, and saves me a lot of time when writing Perl code for SQL operations like this.