Teoria Ewing?

More Questions than Answers with Diamanti’s Performance

When I first sat down to write, I was planning to follow up on my previous article on Serie A relegation battle. I was going to go through the list of teams likely to be in relegation battle and demonstrate what numbers said about their many shortcomings and few bright spots. I started my analysis with Bologna and if not for kindness of Jörg Seidel (owns the very interesting Goal Impact site) who agreed to send me his ratings on Bologna players you would be reading a completely different article right now.

What is Goal Impact? Goal Impact measures the extent to a player contributes to the goal difference per minute of a team. If a player has a high value, then his team had a higher goal difference per minute in games he played versus the games he was absent. Goal Impact is an advanced +/- rating, and like all models it is not perfect but it is universal (can rate players from any league). In basketball analytics which I also follow closely there is a feud between +/- and regression guys as to which approach is better. Famous “Ewing Theory” popularized by Bill Simmons is a qualitative and simple +/- approach at heart. (You can find a rather puritan critique of +/- approach here. )My goal here is not to suggest one approach is better or worse. I find value in both. Below we have an example how Goal Impact challenges some of the more conventional wisdom derived from statistics sites as WhoScored and Squawka.

All the numbers I looked before Goal Impact’s at suggested that Bologna had one star level player and rest was mediocre or worse. The star player is Alessandro Diamanti. Last year he was ninth best player in the entire league at WhoScored and best outfield player in Bologna’s squad at Squawka. Both these sites which base their models on Opta were not the only ones who rated him very high. La Gazzetta dello Sport (GdS) also rated him as the best player in Bologna. (Special thanks to Gab Marcotti for explaining that all major sport dailies send two journalists to top flight games- one to write the main article, and the other just for player ratings, so we can assume a lot of effort goes into this endeavor and is taken seriously).

2012-13 WhoScored Squawka* GdS
Diamanti

7.51

0.33

6.31

Bologna

6.81

0.22

5.8

       
*Per Minute      

In recent years with the help of data collection efforts stats community started going beyond goals and assists.(even tracking assists is relatively new in soccer) Actions that lead to goals such as key passes and shots are increasingly taking the spotlight as important predictors of performance. When I look at this data for Diamanti, it is not difficult to see why he is being rated so high. Numbers suggest Diamanti does everything. He is very high in shots, key passes and dribbles. In top 5 leagues only very few players had more than 2 shots, 2 dribbles and 2 key passes per game:

  2012-13 Top 5 Leagues
  Shots Key Passes Dribbles
Suarez

5.7

2.7

2.9

Cazorla

3

2.5

2.3

Kruse

2.4

2.7

2.7

Ribery

2.4

3.2

4.5

De Bruyne

2.7

2.6

3

Diego

2.1

2.1

4

Diamanti

3.9

2.4

2.9

Taarabt

3.2

2.5

2.7

Models that these sites use find him to be an extremely good player mainly based on the sheer volume of what he does for Bologna. Then I looked at the Goal Impact numbers; what I saw was contradictory. According to GI, Diamanti is the second worst player in Bologna squad. This is quite a drastic divergence between two approaches. Usually both models tend to rate players similarly. Sometimes, the “regression based” ( WhoScored, Squawka) approach will have a bias against good defenders as it is difficult to count actions that do not happen. To simplify a very good defender is not the one who tackles a lot, but someone who does not even have to tackle by superior positioning. Defense is about prevention and offense is about production; first is always more difficult to quantify. A great introduction here.

  WhoScored Squawka* GdS GoalImp
Diamanti

7.51

0.33

6.31

89

Bologna

6.81

0.22

5.8

100

         
*Per Minute        

In offense, we assume more shots and passes are better. Perhaps on team level it is, but in player level there is always an opportunity cost to an action that an attacker takes. If Player A shoots the ball a lot instead of passing he will score more goals in the long term, but might not be maximizing his teams’ goals. Same can be said for passers who do pass on good shots. They are being unselfish but are they taking the most optimal route? We do not want players to maximize their own output but their teams’. In basketball where each possession has a high chance of being converted into points this effect is rather easy to quantify. You do not want a player who just can score 20 points but a player who can score 20 points by using 20 possessions rather than 40. High volume but inefficient scorers are perhaps most overrated players in NBA. Same effect is difficult to quantify in soccer as possession between teams are not equal and number of goals is quite small. But digging deeper in Diamanti’s numbers suggest he might be such a player; a “ball hog”. First let’s compare Diamanti’s final output (goals ex penalties and assists) to his intermediate output (shots, passes, dribbles). I also provide other high volume players’ efficiency metrics and season average for all players in Serie A for 2012-13.

` Absolute Metrics     Efficiency Metrics  
  Shots Key Passes Dribbles   G/S A/KP (G+A)/Dr
Suarez

5.7

2.7

2.9

 

12%

6%

29%

Cazorla

3

2.5

2.3

 

11%

12%

26%

Kruse

2.4

2.7

2.7

 

13%

9%

21%

Ribery

2.4

3.2

4.5

 

15%

16%

20%

DeBruyne

2.7

2.6

3

 

11%

10%

19%

Diego

2.1

2.1

4

 

10%

10%

11%

Diamanti

3.9

2.4

2.9

 

3%

9%

11%

Taarabt

3.2

2.5

2.7

 

5%

5%

11%

               
Avg        

10%

         NA

23%

Numbers suggest he is not getting much results from playing with the ball so much. Perhaps this might be due to him playing with relatively weaker teammates than Cazorla and Suarez. Yet one must add that Bologna is usually considered a team where strikers go to revive their careers. In last 2 years Diamanti played with Di Vaio and Gilardino both prolific and efficient scorers. Perhaps there is a problem with Diamanti’s shot selection. (You can read more about shot selection here , here and here at the excellent StatsBomb site.) Diamanti takes 58% of his shots from outside the box. That is an extremely high figure for a high volume shooter considering only one of his goals were scored from outside in open play. Last year in 5 major leagues only Ronaldo, Negredo and Suarez took more long range shots than him. In 2011-12 Diamanti had Ramirez and Di Vaio with him, so his shot numbers were not as high as last year. In 2012-13 he decided to replicate some of those shots himself.

  2011-12 2012-13 Marginal Shots
Shots

92

132

40

SoG

32

39

7

% on Goal

35%

30%

18%

One can assume certain shots are no-brainers to take; either by location or situation. Some others might have a reasonable expectation to go in. And few more are complete lottery tickets. Players do all of these, some because they are irrational or selfish, some because they have no other option but to shoot (e.g. counterattacks, no help from teammates). Truly great players are not the ones who can shoot or pass well, but know exactly whether to shoot or pass. We can interpret the data in two ways. First is to say Diamanti is relatively good decision maker as when he had other shooters with him he took fewer and better shots. Or we can conclude his decision to replace their shots by himself is a terrible decision as his marginal shots are extremely poor, and he is costing his team.

Another problem with Diamanti’s performance is that he seems to have increased his dribbling and passing without really improving team’s performance. It is one thing to dribble and pass or shoot if you are good at it; Messi and Ribery also dribble a lot. It is a completely different thing to dominate the ball as much as Diamanti does. Many of his dribbles are futile; he was by far the most fouled player in top 5 leagues last season. Perhaps this suggests that he holds on to the ball too much and therefore gets fouled often. Diamanti was also ranked second in dispossession and third in turnovers last year in Serie A. Adding all these up gives us a profile akin to the talented kid at schoolyard who is always dominating the games but not necessarily winning many.

To sum up, in Diamanti we have an example of an attacking player who dribbles, passes and shoots a lot. Many models and experts rate him high. He plays for Italian NT. He might be doing a lot on the field but perhaps he could be more valuable if he did less. Even though more data and analysis is required to quantify the opportunity cost Diamanti gives up by being involved so much, we can assume that he makes Bologna a much more predictable team as everything goes through him. It may be that without him Bologna is a better team as each of the remaining players share his burden and the end result is more successful. To give him benefit of the doubt, he might be better off in a “team of equals” as to see how he performs where he is not focal point of offense. His performance for the Azzurri might require further analysis.

Diamanti is rated pretty high by GdS and Prandelli but for a 30-year-old his resume is quite weak; lots of Serie B in his early 20s, then an unsuccessful stint at West Ham and finally constant relegation battles in Brescia and Bologna. That might be telling us something; his style is not considered suitable for bigger teams due to his dominance of the ball on the pitch. Only player that is really similar to him Taarabt (GI rates him higher @ 98.2) of Fulham. Since EPL is more popular among fans and statistics crowd he can also be analyzed to answer some of the questions Diamanti performance raises:

  • Is more always better in offense?
  • What is true opportunity cost of certain actions that players take? How can we quantify it?
  • What efficiency metrics can we develop from the data available? Some of mine are above.
  • How can we incorporate +/-, regression based and other models into one larger framework?

I hope more people like Mr.Seidel develop their own original approaches to soccer analytics and at the end we have many tools to rate and analyze players and teams rather that a one fits all approach.

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s