Premier League 18–19: A Review By EDA
The 2018–2019 Premier League drew to a close last month but not without providing a roller-coaster (yes, I went there) of emotions. Manchester City’s star-studded squad put on dazzling performance after dazzling performance to ensure they retained the premier league in what was one of the best title races in recent memory.
English clubs were successful in European competitions, with 3 English teams reaching the quarterfinals of the Champions League with Liverpool and Tottenham progressing to the final in Madrid. Arsenal and Chelsea’s progression to the final of the Europa League ensured that the trophy came back to London. Tottenham let down their London counterparts in that regard.
This year’s premier league has some interesting and unexpected findings plus some rather obvious ones presented with a statistical backing.
All data and (pretty much) all code and plots used in this can be found on my Kaggle kernel
Reviewing by Exploratory Data Analysis
Cardiff City, managed by lifelong referee critic and purveyor of inspirational half time talks, Neil Warnock, were relegated alongside Huddersfield Town and Fulham. In doing so the two former teams have taught us its important to find a healthy balance of foreign players or you might find yourself in a spot of bother.
Cardiff City had the least of any team in the premier league with only ten foreign players, Huddersfield Town had the most with 27 but still couldn’t avoid the drop. I’m assuming Nigel Farage would be rooting for Cardiff’s philosophy more so than Huddersfield’s. Top tip : when betting on next year’s relegation candidates look for a team with =< 10 and >= 27 foreign players, with this crafty formula up your sleeve you’ll be guaranteed not to lose as much money as you did this season at the bookies. But since Fulham were also relegated Its probably best not to use that formula.
It would appear that age is linked to success in the premier league (to a small amount).
The top seven ranked clubs have average squad ages between ~27 and ~28. 65% of teams have average squad ages between 26 and 28. It would suggest that such clubs have a good balance of youthful talent and seasoned experience in their ranks. Cardiff have a highly experienced side, with an average age over 29, though that didn’t help them in their bid for premier league survival. Newcastle and Bournemouth have the youngest squads and have shown signs of promise.
Manchester City showed us that spending big will land you silverware, that, and in the cross hairs of UEFA’s Financial Fair Play Police. Manchester City’s net spend since 2010 is greater than up to 10 other of this year’s premier league team’s net spend combined.
City have spent over 1 billion pounds, their next closest rival, both geographically and in net spend, are Manchester United. Yet United’s net spend is almost half that of their city rivals. On the converse, Tottenham are championing the cause for misers, showing that you don’t have to spend big to finish runner’s up in Europe’s biggest competition, having spent 16.78 million. Tottenham’s expenditure is considerably less than feeder club Southampton (33.44 million) who have made a habit of selling their best players to Liverpool. However, in the last ten years Spurs have made huge profits on the sales of current Ballon’d’or winner Luka Modric and the non-Spanish speaking golf sensation that is Gareth Bale.
Since 2010,on five separate occasions Manchester City have spent over 100 million £ in a season. The rest of the league combined have crossed that threshold eight times, three of which were by Manchester United post Sir Alex Ferguson.
Money, Money, Money, Monneyyyy
There is a large gulf between the top 6 and the rest of the league in regards to player value and overall market value. While this value is subjective and transfermarkt.co.uk does not release valuation methods it’s hard to understand the driving factors behind market value. Here is an interesting piece of research regarding player value’s in football. Individual performances and accolades alongside a teams’ final league position will undoubtedly alter squad value and in turn average player value. After the Liverpool’s champions league win it is a certainty that every player has added a few million to their price tag. This year’s top four (City, Liverpool, Chelsea and Spurs) average player value is ~30 million £ or greater. The entire top six’s, which includes the teams above plus Manchester United and Arsenal, average player value is greater than 20 million £. Of the remaining 12 teams only 4 have an average player value of more than 10 million £.
Manchester City have the most expensive squad yet they have not spent over 60 million pounds on a single player to date. Liverpool have a number of expensive stars, namely the world’s most expensive defender, Netherlands Captain, Virgil Van Dijk and the world’s second most expensive goalkeeper, Allisson Becker, who along with his wife Dr Natalia Loewe were recently made UN goodwill ambassadors. Astonishingly, Virgil Van Dijk and Allisson Becker cost more than Fulham and Huddersfield Town’s squads.
Using a Kmeans clustering algorithm with k =2 as ascertained through silhouette method and a visualization using a principal component analysis. The gulf between the top 6 and the rest of the league when comparing net spend is abundantly clear.
Arsenal are placed away from the regular big spenders: the two Manchester clubs, Liverpool, Tottenham and Chelsea. This illustrates the difficulty that less wealthy teams face when looking for survival and title challenge within the premier league.
Shooting, Passing & Possession
Manchester United, despite finishing sixth ranked the best shooting accuracy being the only team with an accuracy over 40%. Unsurprisingly, Liverpool and Manchester City are also near the top, second and third respectively. With players such as Augero, Sterling, Salah and Mane its no surprise they rank well. Surprisingly Bournemouth are number four for accuracy despite being in the lower end of the shots per match table.
Bournemouth ranked top for number of goals scored on the counter attack followed by Leicester and Liverpool. Both Liverpool and Leicester were within the top three the previous season also. However, I’m unclear as to what exactly constitutes a counter attacking goal
When looking at Long passes and Backwards passes Bournemouth also find themselves in an interesting situation. Bournemouth are stylistically contrasting to the teams placed around them in the league table.
When analysing long passes per minute, Manchester City and “Sarri Ball” Chelsea alongside Bournemouth seldom make long passes, however, unlike Manchester City and Chelsea, Bournemouth do not tend the play the ball backwards and like to play direct. Cardiff on the other hand do not recycle the ball backwards and appear to like to hoof it long, reminiscent of football 30 years ago. This may be something to do with Neil Warnock’s anachronistic tactics…and self.
Fulham, despite not fairing well in the league looked to play “football” throughout the campaign. Fulham ranked well for overall possession, as compared to their league place neighbors and also were willing to play the ball back, more so than the league average.
When compared to other teams around them in the league, Bournemouth’s average possession per match fairs well, despite being below 50%. There is a strong correlation (0.79) between possession and backwards passes.
Its hard to uncover what this truly means but there are two schools of thought 1) The longer the team hold the ball the more chance there is that they will be required to pass back 2) or that the more you pass back (less risk of loosing the ball) you retain the ball for longer and ensure a higher possession statistic. The top three for possession and backwards passes are Manchester City, Chelsea and Liverpool.
Newcastle,Cardiff and Burnley make, on average, the most long passes and record some of the lowest backwards passes whilst also having some of the lowest possession in the league. This suggests a direct style of play from these teams as they employ route one football. This is in direct contrast to the higher placing teams whom emphasize retention of possession, even if that means going backwards, punctuating play with a series of short passes.
A means of understanding similarity between teams for a variety of variables is Kmeans Clustering. Some of the methodology was mentioned above, to get a better understanding of the methods used you can click the link.
Using the clustering algorithm we can understand to an extent, stylistic similarities between teams. This was also done above interpreting multiple plots in tandems, with this we can use a principal component analysis to visualise the clustering. This type of clustering has some drawbacks but this isn’t really the time or place to get into it. To fully understand why clusters contain specific teams you will need to have an idea of the data. You can view the data and create your kernel. It would be great to see more people involved.
But for now here are some of the Kmeans Cluster Analyses visualised by principal component analysis.