EPL Prediction Model: Final Day | Liverpool, City & Chelsea to win

EPL Prediction Model: Final Day | Liverpool, City & Chelsea to win

We’ve arrived at the final weekend and it’s been a lot of fun this season. Some small ups and downs at the start, then around the turn of the year the model really started to fly and we were averaging over six out of ten correct calls per week. The past month hasn’t gone well but is providing good motivation for a rebuild of the model this summer. It will definitely be back next season in a new and improved form, so make sure to look out for the 2014/15 predictions!

Two factors have led to that lower recent performance. One has been an issue all season and that’s the fact that these posts are written a couple of days (at least) before the teams are announced, so simulations are based only on a best guess at who’s playing. More automation of the model next season should mean that we can run an instant update on Saturday as the teams are announced and then bet using confirmed starting elevens.

The second factor is that the model as it stands uses a whole season’s worth of data to simulate how players perform. Last season that worked really well, but this season Palace (in particular) have caused all sorts of inaccuracies. The new model will need to understand when a team fundamentally changes performance and adapt to that.

Finally, the bit I’m really excited about is next season’s EPL Model is going to use event location data from Opta for a much more detailed game engine. We won’t just know who passes to who, but where those players are on the pitch, the chances of completing that particular pass and the chance that the player who receives the ball will be in a position from where he can shoot.

Stay tuned on Twitter to @data_monkey and you’ll be the first to know how it’s all progressing as I spend the World Cup writing a new match simulator code…

It’s time for the last set of percentages for 2013/14. And sorry Liverpool fans, you’re expected to beat Newcastle easily but there’s only a 16% chance Manchester City will lose at home.

final day

Here are the simulated passing and shooting stats:

final day - detailed

And the model’s predictions:

Cardiff City v Chelsea – Away win
Fulham v Crystal Palace – Draw
Hull City v Everton – Draw
Liverpool v Newcastle United – Home win
Manchester City v West Ham United – Home win
Norwich City v Arsenal – Away win
Southampton v Manchester United – Away win
Sunderland v Swansea City – Home win
Tottenham Hotspur v Aston Villa – Home win
West Bromwich Albion v Stoke City – Home win