Seems like the proliferation of probabilistic-based election models has slowed down, which makes me both happy (less competition, fewer nerdfights, people don't duplicate one another's work) and sad.
The "sad" part is because I think these models have a lot of value and I think 2016 actually demonstrated that value in a lot of ways, at least for the ones that gave Trump a reasonably good chance. I don't think one should cater to one's critics. But enough about that.
I just want to underscore that building these models is a *lot* of work. You have to have a lot of domain knowledge about politics, data visualization, coding and (especially) applied statistics. There are lots of weird edge cases and if you take shortcuts you tend to pay for it.
On top of that, the data collection process is a huge task unto itself (especially for the House with 435 races). And communicating about them can be challenging. It's a real team effort.
For data-y stuff, it's not always easy to tell what has a lot of thought and effort behind it and what doesn't. In general, though, give the benefit of the doubt to models that "show their work" and can explain what they're doing *and why they're doing it* in plain language.
