Spurred on to 4th?

As the Premier League season enters its final weekend, it is quite unusual that across the league there is very little to play for. The one key issue that remains to be decided is who will finish in the final Champions League position.

Following its success last season, one team that is trying to repeat the feat and establish itself as a regular Top 4 team (at the expense of its North London rivals) is Tottenham. After an inconspicuous start to the season, with a new manager in charge, Spurs have had a pretty good year, with performances often lit up by the PFA and Football Writers Player of the Year, Gareth Bale.

Ahead of this weekend’s final game, Onside Analysis has spent some time evaluating who have been the most important players for Spurs.

Before we dive straight into our results, we need some explanation of our methodology, to help put what comes out into context. If this was a television advert for women’s shampoo – this might be the part they call “the science bit”.

We begin with one of the statistical models that we like to spend our time building at Onside Analysis. Without going into too many details, the model we are using for today’s results is simply a team-based model, which uses historical data to determine the strengths of each team, taking into account the quality of opposition. (For those who like details, it is our extension of the adjusted-Poisson model first introduced by Dixon and Coles, 1997). Importantly, the model itself does not take any account of player actions or even the players who actually played in each match.

Having a statistical model of this type allows us to make probabilistic predictions about matches. In particular, for each game, we can say, before the game began, what were our best estimates of the chances of every possible score-line. This then allows us to define the expected number of goals each team might score. Without getting too geeky here, this is a very specific definition of the word “expectation”; the result of a mathematical formula, which takes into account the probabilities of each possible number of goals. Importantly, it is not necessarily a whole number, but can be thought of as an estimate of the average number of goals a team would score if the same match was played many many times.

Armed with our Expected Goals for both teams in each match, we can compute (for each match) an Expected Goal Difference (the difference between a team’s and its opponents’ Expected Goals), and an Expected Total Goals (the sum of the two team’s Expected Goals). Then, for each game, we can observe what actually happened, and look at the difference between the Actual Values and the Expected Values. A positive value of the difference between Actual and Expected Goal Difference means that the team did better than expected for the particular match, whereas a negative value means they did worse. Similarly, if the difference between the Actual and Expected Total Goals is positive, the teams scored more goals than was expected and if negative they scored fewer. We can then average these differences over all games a team plays in a season, to get an idea of how they have performed compared to how our model expected.

Hopefully, you’re still with me. There’s a little more of the science bit left, but at this point it is probably useful to include a picture.

Clearly there’s a lot going on in this graph, and we haven’t even mentioned players yet, but we will soon. For now, if we look at the top row of the graph, we can see the average performance for Tottenham this season. The bar shows the team’s average value of the difference in Actual and Expected Goals, over the 37 games of the season so far. As we can see, Spurs have actually performed better than our model predicted (by a goal difference of 0.11 goals – the number directly to the side of the bar). The green colour of the bar is also important: it signifies that in Spurs’ games there have been more total goals than expected. If it had been blue it would have indicated fewer goals than expected, and grey would have meant (roughly) as many goals as expected.

So far so good, but our purpose was to look at the players. And this is how we do it. For each player, we proceed similarly to what we have done for the team as a whole, but averaging only over the games in which the player was in the team. By “in the team” we decided that a game gets included for a player if he played 50% (45 minutes) or more of that game.

This leads us nicely to interpreting the rest of the graph. The bars for each player (and the value immediately to the side of the bar) indicate the average difference in Spurs’ Actual and Expected Goal Difference for the matches that the player was in the team. The colour of each player’s bar indicates whether the sum of both teams’ goals was above or below expected for the games that player featured in.

For each player, we can also compute these average differences between Actual and Expected performance, for the games where the player was not in the team. This number is shown in square brackets next to the value for the player when he is in the team.

To end the science part, the numbers directly after each player’s name indicate the number of games that the player has been in the team (according to our definition of “in”), as well as the total percentage of minutes of Spurs’ Premier League season that the player has actually played. Players with fewer than 5 games “in” the team have been excluded from the graph.

If this is all too abstract, it is perhaps helpful to look at an example. We’ll take Gareth Bale (the 6^th bar and 5^th player down). Gareth Bale has been in the team (i.e. played more than 50% of the game) in 32 out of Spur’s 37 fixtures. He has played 85.1% of Spurs’ total Premier League minutes this season. When he has been in the team Spurs have had an average Actual Goal Difference 0.20 higher than Expected. This is notably higher than Spurs’ overall value of 0.11 (which can be easily seen with reference to the vertical dashed line).

If a player were to play (by our definition) all of Spurs’ games, then necessarily his average would be identical to Spurs’ average. Given that Bale has played the majority of games, the fact that his value is notably higher than Spurs’ overall value is evidence of his importance to the team. This is further demonstrated by Spurs’ performance in the games Bale has not played. By looking at the number in brackets next to the bar, we can see that in such games their average Actual Goal Difference has been 0.45 less than Expected.

In our opinion, the ranking that makes most sense, is something similar to the difference between the performance when a player is in the team and when they are out, but taking account of the number of games in and out (so a single result doesn’t have too much weight). To avoid making the graph even more complex, we have neglected to include this value. However, we have computed this measure and this is how we have ordered the players (which is why the bars are a bit higgledy-piggledy – to use a technical phrase).

And this draws us to an interesting point. While Bale is clearly highly ranked, he is not the highest ranked player, with Dembélé coming out on top, and Naughton, Lloris and Sandro also above him.

Before we explore this further, it is time for some caveats. In particular, because this measure only looks at team performance when a player is in or out, and does not directly take account of player performance, we are not saying who the best players are, rather who the most important players to the team are. A player may be important to the team for a number of reasons that have nothing to do with his own ability: the quality of replacement players might be much weaker than that player, or having a player in the team might mean that the team line-up or play in a different way that suits the team better and enhances the performance of other team mates.

Bearing this in mind, looking at the order of the players, we believe that most people who have watched Spurs a lot this season would largely agree with the majority of the players near the top. Interestingly, the decision to sign Lloris from Lyon, in spite of having a previously high-performing (if ageing) goalkeeper looks vindicated. But why does Dembélé come out top (and he does by quite some margin). This is what one of our Performance Analysts thought:

In the system that Tottenham typically use with a 3 man central midfield, it is customary to have a box to box midfielder and for Spurs this season that man has been Mousa Dembélé. There are a lot of teams who are severely weakened in the absence of that type of player from their midfield as the dynamics of the team can be lost if they do not have a suitable replacement. For example Manchester City without Yaya Toure, West Ham without Mohamed Diamé both stand out, whilst Aaron Ramsey has recently started to excel in the role at Arsenal. No other player in the Spurs squad can quite do what Dembélé does for Spurs as he is an all-round midfielder with the only thing lacking from his game at the moment being goals.

 

Dembélé certainly has the ability to break up the play and make challenges and if the opportunity is there, he can take the ball forward for Spurs, turning defence into attack. He glides past players with ease, using an excellent turn of pace and that has been crucial at times in breaking down resilient opponents whose defences have not been breached with either Tottenham’s quick passing or a spectacular goal from Bale.

 

The players around him only possess a handful of his talents. Huddlestone excels as an excellent long-range passer of the ball, Sandro is a good defensive midfielder whilst Parker is busy and industrious. Parker can tackle but is not always aware of his role as a defensive midfielder. He shows great energy to carry the ball forward but does not always have the productivity in the final third. Dembélé is able to do nearly all of these things, although he is not a defensive midfielder and so the absence of Sandro has certainly been noted for Tottenham as you can see by him being the 4^th ranked player.

It is fair to conclude then, that as the numbers suggest, Dembélé’s influence over the course of the season has been significant. The value of a player is often judged by the contribution when they are playing, but it can be just as telling as to what happens when they are not on the pitch.

The plaudits have largely been handed to Gareth Bale this season and rightly so. However, the impact and importance of Dembélé, cannot be underestimated and Spurs are unquestionably a much better side with him in the team.

This article first appeared on Onside Analysis – check out their excellent work here.