By Ted Knutson

[I wrote this intro for my initial predictive model piece, but I’m guessing most people haven’t read that one, so it bears repeating here.]

I work for a very large internet bookie. Because of this, I’m heavily restricted in what I’m allowed to write about.  So despite the fact that I’m a subject matter expert in English Premier League betting, I ain’t writin’ about it, no way, no how. You’ll see me occasionally discuss my opinions on Champions’ League matches or College Football on Twitter, but that’s about it. (I don’t actually work on the Champions’ League product and used to bet College Football professionally before working at Pinny.)

Anyway, I do other stuff at work too, including dabbling in predictive models. I can’t produce weekly picks for you, but one of the things I can do is discuss some point-in-time power rankings. Most of the models I’ve looked at become active after week 5, including the latest iteration of my own. For fun, I’ll post the initial rankings on StatsBomb for most of the European leagues as they come online, and you guys can argue about them until your heart is content or filled with hatred. Either way.

Why Should I Care?
You probably shouldn’t. This model is just stuff I play with during my spare time, and I’m certain it doesn’t have the brainpower or man hours put into it as something like the European Club Index from Infostrada. On the other hand, the most recent version of this model has been in play since June, and has done a good job picking out the good and bad teams in Brazil and MLS very quickly.

Simpler versions of this model also predicted Augsburg would pick themselves up off the bottom of the Bundesliga and escape relegation last season, which they did. Sadly, it also thought Deportivo would save themselves from doom in Spain, but missed that one due to three losses in their last four matches. Ya win some long shots…

A quick note on methodology: These rankings know nothing about the league table. They don’t care what team is in first and what team is in last, and operate on an entirely separate statistical methodology. They are highly predictive of future league places in the table, but variance exists. I would not post them if I didn’t believe they were pretty good.

Additionally, these are very early rankings. Things tend to settle down a lot after week 10, so expect some volatility between now and when we next check back on them.

EPL Rankings After Week 5

1

Tottenham

2

Manchester City

3

Chelsea

4

Everton

5

Arsenal

6

Swansea

7

Southampton

8

Manchester United

9

Hull

10

Liverpool

11

Aston Villa

12

Newcastle United

13

Crystal Palace

14

West Bromwich Albion

15

West Ham

16

Stoke

17

Norwich

18

Cardiff

19

Fulham

20

Sunderland

 

Predicted League Champ: Tottenham
Fighting for CL slots: City, Chelsea, Everton, Arsenal, Swansea, Southampton (United – See Below)
Relegation candidates: Cardiff, Fulham, Sunderland

Before you start spewing hate at myself and the model, a few things to note. These are not mathematically adjusted for schedule strength. Manchester United have had a very difficult start to the season. Everton have had a fairly easy one. Thus there is strength of schedule skew involved, and while I think I’ve solved some early season SOS issues, the solutions are not ready for primetime.

Additionally, teams will move through the course of the year. Sunderland were dead bottom for about 30 matches last season, but drug themselves out just in time to not get relegated. Last year, the bottom 3 in EPL according to the model were West Ham, QPR, Reading. In Bundesliga they were Hannover, Fortuna Dusseldorf, and Greuther Furth. It’s not perfect, but it gets pretty close.

Arsenal and Swansea are where they are by virtue of redonkadonk shooting percentages. For Arsenal, this is known as, “The Mesut Ozil Offect.” For Swansea it’s known as, “Michu ain’t regressing yet.” Probability says both will cool down a bit as the season goes on.

Probability has also been known to lie.

Sunderland
I mentioned how Sunderland were really quite bad most of last season. This season, they were so bad they broke the model. Basically they did everything poorly except for the shots battle, and Crazy Paolo had no idea what he was doing. He’s gone now.

What does that mean for the rest of Sunderland’s season? I have no idea. No one does. I don’t think this team is as bad as their coefficient looks but… Thus far they have been putrid at doing things that matter. That could change quickly. Or it could stay that way the entire year.

It may take quite a while for even a good manager to figure out what he has and how to play them. Their upcoming run of games mean things won’t look much better for some time.

Cardiff
Schedule issues make them look like relegation candidates (Everton, City, Spurs is a tough three out of five). I think they are better than this, but my eyes don’t always know the whole story. The next five games will go a long way toward telling us where they will likely be at the end of the season.

A note for both Cardiff and Fulham: Only creating 8 shots a match will get you relegated. Period. In fact, no team in the last five years of the Premier League has posted fewer than 10. You’re going to need to step it up.

Tottenham
Andros-TownsendAt the top, Spurs have the best coefficient by some distance. On the other hand, they have had some easy matchups in Crystal Palace, Cardiff, and at White Hart Lane against Norwich and still have trouble scoring. Swapping in Lamela for Andros Townsend will increase their goal expectation considerably.

Aside: Townsend is terrible at choosing where and when to shoot the ball. He has 22 shots already this season and zero goals, and that’s not really a fluke. He has been a stone cold killer of attacks so far this year, and someone needs to sit him down and coach him out of it, fast. “Repeat after me, Andros: You are not Gareth Bale.”

Do I think they will actually win the league? *Kicks some pebbles and shrugs* Probably not? But I feel pretty strongly they will take a Champions’ League spot this season and the model does too, which means someone important is getting pushed out. The gnashing of teeth when that happens will be epic.

Liverpool
I wrote about this in detail yesterday[LINK]. “I gotta ask, Brendan… where’s the tiki-taka?”

  • Adrian

    Spurs in CL next season….I really like you Ted, even if you are a Gooner.

  • Cooper

    Thanks Ted, interesting article.

    You say that rankings are ‘highly predictive of future league places in the table’ but how far in the future are you predicting and how have proven this apparent ‘high’ accuracy?

  • Derek

    Why don’t you adjust for schedule early? That seems to be the most crucial time to adjust seeing as teams have different amounts of home games and wildly different schedules. I do one of my own with a schedule-adjustment and City take a hit (Hull, Cardiff) while Man U are top 4.

    This caused the most problems in your Bundesliga model where good teams were being put in last (Frankfurt) simply for playing Dortmund and Bayern. In my model, Frankfurt was 6th after 5 games. I’m not saying mine is perfect or anything, it’s pretty basic stuff but you have to adjust for schedule or you get results that go against common sense like Frankfurt last and Braunschweig not in the relegation spots

    • tknutso

      There’s a LARGE difference between knowing that there are problems with strength of schedule and mathematically adjusting for it. As I mentioned, I have a couple of solutions for SOS adjustments, but nothing I am super happy with at this time, and thus I left the rankings as is for now, but noted the issues.

      As for “common sense…” that’s exactly what models are for. There have been a ton of times in the last year where common sense says X, the models say Y, and the models are consistently correct. Correcting with common sense is bad, mmmkay?

      • Derek

        didn’t mean it to be harsh, just seems that after say 18 games it’d be ok to not adjust for strength of schedule but right now models aren’t that usable if you aren’t adjusting for teams played. by common sense I didn’t mean eyeballing it just that if teams that play Bayern and Dortmund wind up on the bottom consistently while teams who’ve played Bremen and Braunschweig wind up higher something needs to be changed. comparing a teams actual shot attempts for/allowed to expected shot attempts against their schedule is what I do.

        I love the site, one of my favorite of the kind as I’m a former baseball stat nerd who has just gotten into soccer in the past couple years and am just really looking into the statistical side. so far to go really, there are opportunities everywhere.

  • http://fantasyformation.com Gummi

    I just wanted to pop in and say thanks for all your writing.

    Regarding schedule, I don’t know how you could adjust for it at all. If you do it early in the season, you either have to rely on last year’s schedule, which are different teams, both in relegation/promotion and new players, or go with common sense, which completely goes against the purpose of the model.