EPL Forecast Comparison

Quite a few people are producing probabilistic forecasts for the outcome of the EPL this season and so I thought it would be interesting to collect and track them as the season progresses. Specifically, I’m monitoring the probabilities assigned to the contenders for finishing the season as:
  1. champions;
  2. in the top four (champions league places);
  3. in the bottom three (relegation positions).
The goal is to keep a record of the forecasts made following each round of EPL games, commencing on the completion of gameweek 22. While this is ultimately just for fun, I’m curious as to what drives the spread in forecasts and how they fluctuate over time as different methodologies apply different weights to new information. I’ll write up some conclusions in a blog post at the end of the season.

To properly evaluate forecasting skill, I think you’d need to collect and assess match outcome predictions. It would be very interesting to set up a controlled experiment. That’ll take some thought: something to discuss over the long barren summer.

For now, I’ll collect up predictions from those that want to participate and present them here, along with some summary analysis. How the results are presented will undoubtable vary over the next 16 weeks. It’s a community project so I'm happy to take suggestions.

I’ll put out a twitter request for updates each week (@eightyfivepoint). To submit, either alert me on twitter, or send an email. Data will be made available to anyone on request. It would be great if participants could also provide a brief write-up of their methodology at some point (data used, and how) to help classify and interpret the results.

Here are the forecasts collected so far for Gameweek 25. I (still) promise I'll find a better way of providing them.

All Forecasts (click on table to expand)


Popular posts from this blog

Freedom of Movement of Footballers: Brexit and the EPL

Consistent or Calamitous: How good is your goalie?

Bodies on the Line: Quantifying how defenders affect chances.