Probably, though my brief literature search didn't turn up any examples of people actually measuring game difficulty using Fitts' law. One thing to keep in mind is that motor movements are only a portion of what might make a game physically or cognitively demanding, even in platform games which are typically quite simple. Users have to make additional decisions, such as which platform to jump to or where to attend (do I look at the platform I'm jumping to, or do I look ahead to anticipate the next jump?)
I did find one interesting paper by Looser, Cockburn, & Savage (2005), who use Fitts' law to predict movement time in first-person shooters. The authors found that the time taken to pan (using a mouse) to acquire a target could be accounted for by Fitts' law. So if the time to acquire a target (e.g. before an enemy starts shooting at you) is critical, it certainly suggests that your methodology could be used to moderate difficulty level, say, by placing enemies farthest away from the current position.
Again, this is mostly speculation as I don't think there are any actual studies that address your specific question. But your intuition seems quite reasonable.
Looser, J., Cockburn, A., & Savage, J. (2005). On the validity of
using First-Person Shooters for Fitts' law studies. Proceedings of.
People and Computers XIX, 2, 33-36. PDF