I have my B.S. in Applied Mathematics and Statistics and currently pursuing a masters in Predictive Analytics.

I believe 2018'will be slower and end around 4-5-% and after 2018, depending on what happens with the overall economy because many predict a "Depression Coming" in 2018, the RE market could see flattening or a loss in appreciation due to an overall economic slow down.

Interest rates below 4%, for most of 2017, and predicted to rise, as Feds plan more increases.3.

(It was me who guesstimated it down to a "95% Refi" assumption).However, have you done a "poor job" with your "$200/m cash flow per door" prediction?

The maintenance can be another variable that can be tough to predict. 

Without having a validly executed incorporation of your business and placing those properties into a business, there are very few ways to protect your assets with any degree of predictability.

The numbers may make sense but usually they are in high turn over areas, high maintenance areas, not so great neighbors and the predictions turn out much lower because of the costs involved to maintain them.The three best cash flow areas I know and watch is areas in Tennessee,Indiana and Ohio.

The hardest thing to do when analyzing a vacation rental is predicting the income. 

When my wife and I talked about moving there, our friend who did flood modeling for the army corp of enginners strongly suggested not moving there—the flooding from Harvey should not be considered a freak occurrence but rather something that will occur regularly (every 25 years or so).I am not making a prediction just some additional food for thought.

Where modulars win are in three categories:  harsh weather environments (which we don't have in the mid-Atlantic), design fees (by reusing the same design over and over again), and cost predictability (a normal builder will price a new design a bit higher to hedge against the unknown).