There’s a few things I want to point out about this, someone is bound to notice them anyway:
Firstly the points total is high, at 1059. In other words it only predicts 81 draws all season, the average is about 100 so there’s probably a 10-40 points discrepancy throughout the league. That’s the nature of this analysis and I’m not going to adjust for it, I want it to be as simple as possible.
In that vein the goal difference is also off, being +19 across the league. Again, I’m not going to correct (or try to) for this, that’s not the point of the model.
The spread of points and goal difference is pretty wide (46 points and 68 goals between first and 20th), but in reality the spread will be wider (an average of 60 points and 96 goals over the past 5 seasons). Random things do happen.
This projections has the table split into four tiers, The top 5, United, 7th – 16th and then 17th-20th. How realistic do I think this table is? Well, with the exception of United, I wouldn’t be surprised to see any of the teams finish where they’re projected so it passes my sniff test. I certainly like the way it spreads out the points compared to the shots models. Are all of the projections going to be right? Of course not, but at worst it’ll be a nice little experiment.