1. No category

We maximize the expected cummulated reward