"Millions"? From your lips, etc....I thought, "man, he goes from having prepared to describe live action, as a PBP guy, to having to just make up whatever about whatever to fill an indeterminate space of time as his production team lines things up". And while I imagine that sort of extemporaneous vamping kinda has to be in the toolkit of any commentator, regardless of job description, I can also imagine where that's a task that stresses you way the heck out, rather than giving you the joy you get from calling a live game that you've prepared for. Ya know, in front of millions of people sitting at home.
But seriously, there's something of a false dichotomy in the way you've phrased this. All play-by-play commentary is improv, apart from any scripted opening I might draft. And on the other hand, a lot of presenting/hosting is the opposite: particularly for the pregame show, but also for parts of halftime and postgame, I know exactly how the segments connect with each other and I can write scripts or at least prepare notes to help me link them together. And even the ad-libby stuff usually fits within a structure, just as play-by-play improv fits within the structure of a sport with normal rhythms and rules. The roles are different, but they're not so different that the difference worries me, particularly as (in my opinion) presenting/hosting is much less of a specialist skill to acquire than play-by-play is. For me, it's a bit like moving down the defensive positional spectrum in baseball: you can move from SS to 3B or 2B a lot easier than you can move to SS from elsewhere in the infield. Play-by-play involves much more knowing when to speak and when not to speak, when to get excited and when to remain calm, when to trust your audience and when to explain stuff it might not understand, and so on...and those elements can vary greatly from sport to sport. Presenting isn't easy, per se, but the skills required to do it well are much more generic and I think more easily acquired, particularly if you're not required to be on camera. (Which is quite an important qualifier, of course.)