Very interesting read. I posted something similar last year (https://towardsdatascience.com/what-the-data-gets-wrong-about-monopoly-8fe8a6941c4e). The main difference is that I use Python to determine landing probabilities analytically (by matrix multiplication), while this uses a Monte Carlo simulation. It is reassuring that the results are similar.

However, I disagree with your emphasis on the number of turns to break even as the most important metric by which properties can be evaluated. I go into this in more detail in my article, but the TLDR is that high revenue per roll properties increase your ability to bankrupt other players (i.e., the purpose of the game). Also, players rarely own more than one monopoly in the crucial phase of the game, meaning that the opportunity cost of e.g. upgrading a green property to hotel is very low.