Best thing to do is probably do an intense refresher of statistics
Then it's just data+SQL and the usual datawarehouse/datamart/cubes/ETL/visualization whathaveu stuff.
For pet projects to learn you can usually find decently sized datasamples of sports and so forth and experiment with that. But anything coming even close to what you'd work on in the real world won't be easy to find.
If all else fails you can always create TBs of fairly random (or more intelligent) data via a script