The Worl Cup is almost upon us and with that, it'll be a good break for me from updating these posts. However, my goal is to publish a review of the first half of the season season just before the Premier League resumes, so stay tuned!
It's amazing to see how much has changed since I initially posted the launch of my Premier League model.
Before the season started, my model had Arsenal with less than a one percent chance of lifting the title, and the Gunners have now increased that figure to an astounding 11% – which, in my humble opinion, is lower than I would've expected given their amazing performances. For context, 538 has them at a 22% chance.
Other major changes have been in Brighton's and Newcastle's expected points come the end of the season, going from 46 expected points to 54 and 58, respectively.
I still think City will lift the title, but the race for the top four will get quite interesting and my hunch tells me that we'll see Spurs and Manchester United clenching the remaining spots, with Newcastle and Liverpool missing out by just a couple of points.
The naive predictions are computed using historical data only, and it's the forecast system that I use to simulate the end-of-season probabilities for the Premier League.
The following table showcases each game's forecasts under this model.
In this section, I present my biased forecasts, which, are based on both my hunch and historical data. For each match, I'll present a detailed viz with probabilities for different events within the match and a small write-up.
The following table shows the summary of my biased forecasts.
Detailed fixture forecasts
How did we do in the last gameweek?
It all seems like we'll be going to the end of the season with a 3% higher error on our model vs. FiveThirtyEight (unless something really unusual happens in the coming months).
Honestly, I had hoped that we might beat them, but I feel like it's been a success to come this close and to provide accurate predictions on the season so far.
I believe that come the next season, there's a good chance that I can improve certain aspects of the model to beat them next time.
Hope you enjoyed today's issue!
If you like my work, please help me by subscribing to my website and sharing my posts. If you get consistent value out of the site, then consider becoming a paying member – this will help me generate more and better content, week-in and week-out, plus grant you some additional perks 😉.
Catch you later!