External tables: Why create statistics?

The documentation for Polybase says that creating statistics for External Tables will help getting better execution plans. This question is about when the External Data Source is another SQL Server. I guess the better execution plan is on the External Data Source, i.e. for pushdown computations (for where clauses, joins and aggregates). To get a good execution plan the optimizer also has to know what indexes are on the table. Can the optimizer use the statistics on the remote server (the External Data Source)? If yes, what are the benefits of creating the statistics locally on the External Table? If the optimizer has too look at the indexes on the remote server, then it might as well look at the statistics... Please provide references in the answer.