Just wanted to leave this here in order to mark down the end of the year projections from my Four Factors regression compared to 538's current projections.
We're about 17 games in so team's overall net rating is "stable" as we say, meaning it's more than 50% predictive of net rating from here on out.
The four factors regression uses priors that are informed by the team's performance in that measure the previous year. The strength of influence of last year's performance is weighted based on how well that factor has tended to carryover year to year.
In general the Four Factor model is more regressed than the 538 projection. The 538 model is also still much closer to its preseason projections. That's evident in the gap between the Houston projections where 538 is considerably higher, and the Sacramento projection where they are much lower.
I would think that the 538 model would do better than the Four Factor regression. It's based on more information, for one thing. But, I am interested to see how close it is and if there is a specific pattern in better or worse projections.
Without further ado, here's the Four Factors regression vs 538.
