What is 'the problem of the direction of time'?

This paper searches for an explicit expression of the so-called problem of the direction of time. I argue that the traditional version of the problem is an artifact of a mistaken view in the foundations of statistical mechanics, and that to the degree it is a problem, it is really one general to all the special sciences. I then search the residue of the traditional problem for any remaining difficulty particular to time's arrow and find that there is a special puzzle for some types of scientific realist.