Low-Hanging Fruit vs. High Risk

Among my various misunderstandings of Agility is the algorithm for deciding what story/task/monster to attack first.  It's either:

Go after the low-hanging fruit, get the process rolling with an easy win, warm up your tools and environment.

or

Jump into the biggest unknown first to reduce the uncertainty, start requesting external resources or services that you might end up waiting on, start researching something you may have a serious learning curve on.

With Sudoku, it's easy. I always go after the low hanging fruit - anything else would be absurd. Every cell filled in adds one more constraint to each of three cross-cutting sets - it lowers other fruit.  There are no externals and no research.

With my linguistic ... uh ... hobby, it's radically different. I want to jump in and BDD me up some Rails, but I have so much linguistics to catch up on...  The low hanging Rspec fruit is tempting, but biting into it won't bring me that instant knowledge of Good and Evil linguistic theories I need.