I haven't found predictive code completion to be good anywhere for Swift, especially anything to do with Apple frameworks.

I've attributed it to the general lack of open source projects to train off.