It is that time again. The month before World Cup. Kids (and adults) are opening Panini packs with excitement to see if they finally got that elusive reserve goalkeeper from Honduras or Algeria. TVs are showing Maradona’s goals against England everyday on million different contexts. National teams coaches are finally announcing their squads and bloggers are outraged at the inclusion of this and exclusion of that.
Our friend Mr.Goalimpact is running a unique project for the World Cup. We volunteered for Italy and received a list of highest GI players as well as others who played for Prandelli recently. We, at dingostats, are obsessive collector of lists. So we did not stop with GI list. We went WhoScored and complied their top players list. We did the same with newspaper ratings. We found the rating for players from 3 major Italian sports newspapers. Finally we compiled salary and transfer value data from various sources. Why all the hassle?
Because they all contain different and perhaps valuable information about Italian players.
For tactics; considering most of Italian defense is based on Juventus and the tradition of 3-5-2 is still alive and kicking in Italy I decided to go with that for my formation. Below are each metric and their top 11’s. Then correlations of each metric and finally a model. Italy’s 30 men squad here. (Great headline!)
This is a top down metric. You can read the creator’s FAQ here. What Goalimpact can capture that others can’t would be a player’s overall impact on team’s results controlling for other players on the pitch. If your team does well when you are on the pitch your score goes up; if not it goes down. You might be a great defender and never touch the ball due to your positioning; top down metrics would capture your value; bottoms up metrics like dingoR less so. What separates GI from others is that it is a lifetime number adjusted for age. Other metrics are seasonal.
Not too many surprises here. Juventus dominates as it should. Except for Pazzini all others made Prandelli’s team. GI matched conventional wisdom pretty well. But once we go beyond the top 11 there were a few surprises in Goalimpact best players list, especially Felice Vecchione of SG Sonnenhof Großaspach in lower divisions(can’t even figure out which) of German football. He might be good but I doubt Prandelli ever heard of him. Well if he becomes famous one day you heard him here first thanks to Goalimpact.
A completely different approach. WhoScored rates players by their actions on the field. You do something “good”, like scoring a goal or intercepting the ball your score goes up. You get thrown out of the game it goes down. It has some drawbacks like all metrics. Most important of all, it is not per minute; and therefore punishes substitutes. To overcome that I took WS ratings of players in games they started. So who would WS have picked:
Here the list is a bit different from Prandelli’s 30. Consigli, Atalanta’s goalkeeper is not one of 3+1 keepers called . Prandelli is a bit of a fitness fanatic, so both Lucarelli and Totti who are older than me had no chance. Luca Antei was good but he is young and his team conceded 70 goals this year. Managers do not work in a vacuum; they have career risk if they sway too much from conventional wisdom. Diamanti is an old friend of this blog, and let’s just say Prandelli did well by not including him.
Before we could watch every game on TV and get the stats in real time there were newspaper ratings. In a sport where fans barely get any meaningful stats they were(and perhaps still are) the most important numbers associated with players(except for goals) . Italian newspapers always took this task very seriously; sending a dedicated person to each game just for ratings. To us; these area very good proxy for scouts and also informed public. It also reveals some preconceptions about players; a well liked veteran(Totti) might be getting higher grades but an unconventional player(Balotelli) might be getting punished due to his antics. I also think players are rated versus expectations as it is only human to do so. We took 3 main Italian sports papers and used their season ratings for the below table.
Totti was the highest rated player in Serie A by newspapers; a little sentimental. Toni and Bovo are also on the wrong side of 30. Romulo(Brazilian Italian) has been included in provisional 30 which is perhaps due to the high praise he got all year in media. Balotelli was one of the lowest rated in our whole list which is more of a reflection of his character and performance versus the expectations.He was not good this year but not terrible either.
This is a composite metric I created using salary data from Gazzetta and player values fromTransferMarkt. Player values punish older players and are just estimates; salaries are real but they tend to undervalue up and coming players(until their next contract). So to come up with a composite metric I used 3* salary+ TM. Well market is market. It is driven by supply and demand but of course player market is not a liquid one. You can’t just buy a Buffon if you wanted to. He is also worth more to Juve than other teams given his loyalty and history with the club. Given all its imperfections, this is an important part of soccer so we should look at what it says.
All 11 players are on Prandelli’s list. Pirlo is missing due to low TM value and his high salary was not enough. And I am not sure if Marchisio’s agent bribed someone at TM, his value is too high. In case you were wondering De Rossi is the highest paid player in Serie A with 6.5M Euros. Parolo the Parma midfielder who might be his teammate in Brazil makes ~1/20th of that. Mr.Parolo, please call Piketty
Adding All Up
What would be a composite starting 11:
|De Rossi (4)|
Not a bad squad. Rossi is a doubt as he is just recovering from injury but has been included in provisional 30. Only omitted player who has appeared multiple times on the lists was Totti and and he is not a Prandelli favorite. Let’s get statistical. Correlations between metrics including whether a player was selected:
I will leave you to interpret the above table.
As a final note I also modeled Prandelli’s selection criteria using the above metrics; and accordingly De Sciglio, Pasqual, Darmian, Parolo, Romulo, Immobile(depending on Rossi’s fitness) and Cassano will be dropped from 30 for the final 23. We’ll see…