

Discover more from Cannon Stats
First Projections for the 2022-23 Premier League
Breaking out the projection model to see what it spits out
The new season is imminent and that means that it is time to run the simulation for how things will look in May.
For this season I haven't really done any changes to the model because I was generally happy with how it performed and I didn't see any obvious tweaks that I could implement in my limited time this summer, so I will roll with this again. Later in this article, I will go through and do a little bit of a retrospective on how things performed last year.
Let's take a look at what we are dealing with to start the season:
I don't think that there are any massive surprises for this initial list.
Manchester City and Liverpool with the title fight. Chelsea, Spurs, and Arsenal with United having an outside shot for Champions League. A healthy midtable with a bunch of teams all about the same where depending on the breaks a lot can happen. The lower table features the teams that struggled last year plus Fulham, and then you have Bournemouth and Forest who are unfortunately looking at a tough season.
A new graphic I made for this takes the team ratings and puts them on the same scale. I think this really helps to see the team tiers and distances between the teams.
I wouldn't quibble too much with where teams are, the biggest change might be dropping Everton down a bit but they still ended last year as the 14th best team even though they ended up in the Relegation scrap.
One of the big deviations from the betting markets right now is that I have Arsenal above Manchester United. I think some of this is really just down to United always seem to have more money going to them making their odds shorter than where maybe the "true" odds would be.
I also have Everton at 1 in 20 to be relegated against the betting markets that have them closer to 1 in 5 to be relegated. For this I think I would split the difference some starting with them at around 10 percent to be relegated.
The next graphic is the one that I think helps the most at this point of the season, showing the distribution of points totals over the course of the 10,000 simulations.
This is another one that helps to illustrate the tiers for the teams and how much variability there still is this season. The last image is similar but shows the probability for the team to finish in each spot.
The season ahead looks like it will be exciting. One of the big things that I have no idea how to model is that we will have a break right in the middle of things for the World Cup, compressing things before and after to be able to squeeze it in.
I don't know if this will help the teams that don't have as many international players as they will be able to start the second half with a six week break to prepare or if this is something that will hurt them more because there will be more matches on short rest than normal.
How the model performed last season
This was the simulation at the start of last season. Below is the difference in points from the initial projection and the final points total.
Overall looking at how the match by match projections went things looked pretty good to me. The model predicted a Home win 41.1% of the time and that happened 42.9% of the time, I predicted an away win 35.2% of the time and that happened 33.9% of the time, and I predicted a draw 23.7% of the time and that happened 23.2%.
Looking at things by prediction bucket looks like this:
Things tracked pretty well staying pretty close to having the actual outcomes match the predicted outcomes. The biggest weirdness came in the 60-65% bucket where I predicted this outcome 24 times and it happened 20 times, while the 65-70% range was predicted 23 times and hit just 15 times. I think that is just a bit of a fluke but I might watch that going forward if the trend continues.
Anyway, I think my takeaway is that this model is good enough, it isn't going to go win me money gambling based on it (it is not designed to either) but I think it does a good job giving an idea for where the season might go.