clock menu more-arrow no yes mobile

Filed under:

Premier League projections, from the winners to the relegated clubs

I have brand new Premier League projections, built off a totally new system. It's fun!

Jamie McDonald

The Premier League projections are back. The system has been fully renoobulated and is ready to go for the new season. Nerds are welcome to skip down to the extensive methodology section to see what went into these numbers.

My analysis suggests we may be looking at another three-team title race. No system worth your time will rate Manchester City and Chelsea outside the top three, or perhaps outside the top two. But based on expected goals, Liverpool deserve a seat at the title contenders table too. Arsenal have an outside shot, and Tottenham Hotspur and Manchester United are also teams in the Premier League.

Title_odds910

My numbers last year consistently rated Liverpool among the top teams, and Liverpool's title contending run last year probably counts as the biggest "get" for expected goals as a projection system. So far this year, the Reds have totaled six points against the most difficult schedule in the league. They were home to Southampton, who look like a solid top-8 side, and then away to defending champions Manchester City and of course Spurs. While the loss of Luis Suarez will surely hurt, so far Liverpool's performances do not raise any statistical warnings. Even that loss at the Etihad was mostly driven by the quality of City's finishing rather than a clear difference in the kinds of chances created.

How Does It Work?

The system is based on expected goals. This is a method which estimates the quality of a team's overall chances created and conceded. For every shot attempted, I calculate an estimated expected goals value based on the following factors:

  • Shot location. From where on the pitch was the shot attempted?
  • Shot type. Was it taken with the foot or the head? Was the shot a direct free kick?
  • Assist type. Was it assisted by a cross, a through-ball or a regular pass? Was the shot taken off a rebound?
  • Speed of attack. How much ground did the attacking team cover, and how quickly, before attempting the shot?
  • Long balls. Did the attacking move include any very long passes?
  • Dribble. Did the shot immediately follow a successful dribble by the attacking player? Did this dribble beat the keeper or an outfield defender?
  • Set play. Was the shot attempted off a set play or from open play?

I'll explain in detail how each of these factors is incorporated down in that methodology section. Close and central locations are good, shots taken with feet are good, direct free kicks are good, dribbles and through-balls are very good, crosses are not so good. Fast attacks are good, long balls are less good, set plays are a little bit bad.

I create team ratings by summing their expected goals numbers for and against from the past three years. I also include team wage bill as a factor, as wages have shown strong utility in estimating team quality. The current system is about two-thirds 2013-2014 expected goals, and the other third a mix of 2014-2015 expected goals, 2012-2013 expected goals, and wage bill. Every week, I will increase the weight on the 2014-2015 data as our sample size expands.

Queens Park Rangers project as a little less likely to be relegated than their promoted brethren Burnley and Leicester City. This is because of QPR's wage bill, estimated around double those of the other two sides. This method could easily fail in this particular case, as QPR have horrifically underperformed their wage bill in the last three years. But overall, it looks to me like including wages helps with the projections.

I project matches using the same method I used last year, simulating the season 500,000 times, match by match, and collecting the results.

Premier League Projections

Because of rounding, the numbers may not add up quite right.

Team W D L Pts GD %Title %Top4 %5th-6th %Rel
Manchester City 23.8 8.0 6.2 79 +41 40% 95% 4% 0%
Chelsea 23.2 8.3 6.5 78 +38 32% 92% 6% 0%
Liverpool 22.6 7.9 7.5 76 +35 22% 88% 9% 0%
Arsenal 18.6 11.0 8.4 67 +21 4% 54% 28% 0%
Tottenham Hotspur 17.1 9.3 11.6 61 +11 1% 24% 33% 0%
Manchester United 15.9 11.4 10.7 59 +12 1% 20% 32% 0%
Southampton 15.0 11.0 12.0 56 +7 0% 11% 25% 1%
Everton 14.0 11.6 12.3 54 +2 0% 7% 19% 2%
Newcastle United 12.3 11.2 14.5 48 -5 0% 2% 8% 7%
West Bromwich Albion 11.9 12.2 13.8 48 -5 0% 2% 8% 7%
Stoke City 12.0 11.1 14.8 47 -6 0% 2% 7% 8%
Aston Villa 11.9 11.3 14.8 47 -8 0% 1% 6% 8%
Swansea City 12.1 10.0 15.9 46 -9 0% 1% 5% 9%
Sunderland 10.1 11.6 16.3 42 -13 0% 0% 2% 21%
West Ham United 10.9 9.1 18.0 42 -15 0% 0% 2% 22%
Queens Park Rangers 9.6 11.1 17.3 40 -19 0% 0% 1% 30%
Hull City 9.2 10.6 18.2 38 -19 0% 0% 1% 37%
Crystal Palace 8.7 11.2 18.1 37 -19 0% 0% 1% 42%
Leicester City 8.5 11.1 18.4 36 -22 0% 0% 0% 46%
Burnley 7.8 10.2 20.1 34 -26 0% 0% 0% 61%

  • The system isn't selling on Manchester United yet. There is no defense for how they've played over the first three matches, given the EPL's easiest schedule so far. But that's only three matches. United had a solid 0.550ish expected goals ratio last year and sport a wage bill rising into the stratosphere. While Louis van Gaal's side are underdogs for a top four finish, they're not out of it yet. My numbers cannot account for the effect of Angel Di Maria and Radamel Falcao, but by incorporating a wage bill adjustment, I give United a boost based on their ability to sign players of that caliber.
  • Not too much say about Spurs yet. The early season numbers are blah, as the stompification by Liverpool at the Lane outweighs the good of the 4-0 against QPR. Last season's numbers place Spurs in a scrum with United, Everton and Southampton for the 5th-8th positions in the EPL, and they project just slightly at the top of that group now. We'll learn a lot in the next month, but it seems likely that another Europa League qualification season is in the cards.
  • Southampton, like Liverpool, could reflect a failure of projection systems so early in the year. Having remade their roster and lost their manager, should we use Southampton's strong 2013-2014 season as the largest part of their projection? Maybe not. But so far the Saints have hung with Liverpool at Anfield, played to a stalemate against a good West Brom side and dominated West Ham. I don't know that I'd rate Southampton as 1-in-10 shots to finish fourth, but I think this is a top half side.
  • Swansea are kind of funny. I was all ready to talk about how I'd overrated Swansea last year because I didn't include an adjustment for speed of attack, and no one attacks slower than Swansea. I had them rated at the top of the mid-table all year last year, but in my new numbers Swansea's 2013-2014 is down in the muck with Sunderland and Villa. Of course, then they went and instead took nine points from three matches. Swansea have been carried by some excellent finishing--did you see that Routledge volley?--but nine early points may be enough to keep them safe.
  • It's too early to say much about the bottom of the table. Burnley is looking like the immediate-relegation side their budget pegs them as, but even so no one is much more than a 50-50 shot at being sent back down to the Championship. Villa and Swansea have extricated themselves from the bottom with hot starts but could easily drop back down if a few results do not go their way. Projecting the rest of the bottom of the table feels like a mug's game at this point. I would be surprised to see Newcastle or West Brom still in the bottom at the end of the year, but beyond that it is hard to say.

Ok. Those are the numbers. If you want to see how the sausage done got made, click on "Methodology" to reveal the nerdery hidden below.

Methodology +

All data provided by Opta unless otherwise noted.

Michael Caley's writing can be found at Cartilage Free Captain and the Washington Post Fancy Stats blog.