Who should succeed Amrabat at Fiorentina?

First things first

I’ve chosen to find who should be next for Fiorentina after Amrabat for two reasons:

• I’m Italian and follow, for obvious reasons, mostly Italian football. I also collaborate with a page about Fiorentina now and then, so I follow Fiorentina close enough to know something about them, and this piece has been in my mind for some months.

• After some time of preparation and building up the tools needed to do this, I’ve also found time between lectures at University to write this.

Why Amrabat?

Amrabat was in demand in January already, he has made clear that he wants to go to Barcelona, so it’s probable that he will be gone in the Summer.

Fiorentina found a good one in that role and position after not renewing Torreira’s loan in the summer. For those who don’t know Torreira, he was probably the MVP for the Florence-based team last season, so they’ve more or less coped well, but it’s not a guarantee this time.

With only 3/4 months before the Transfer Window, Fiorentina needs to have a clear idea of who they want to go for, but let’s suppose they don’t know.

Boring things: methodology.

We need to talk about how I created what you are about to read. It’s mandatory, but if you don’t feel like reading it, skip this part.

PPP

PPP is the name of the visualization and I used it as the foundation of my work to visualize the team playstyle. It’s all based on possession-adjusted metrics and percentages, since it’s focused on playstyle, it doesn’t consider quality, and it develops like this:

Oh, look. It’s Fiorentina’s PPP for this season. What a coincidence.

UMAP and Clustering (part 1)

What I did next, was to use UMAP in order to reduce all the PPP metrics in a 2D scatter graph which presented all the teams for the 21/22 and 22/23 seasons so far from the European Big 5 Leagues, Eredivisie, Primera Liga, and the Mexican League, plus the 22 seasons of MLS and Brazilian League. (All data comes from FBREF, a great site)

Then, I’ve clustered those teams to highlight similar play styles and have every team labeled inside my data frame, so I could easily access them while filtering out things.

This was the result:

Fiorentina is inside the 0 clusters in both seasons taken into account.

So, when I filter players from our player database, I have to look out for players inside cluster 0. I then decided to use clusters 7 and 2, because in many aspects, they’re teams similar to Fiorentina, and they are ball dominant; while teams in clusters 6 and 4 are more pressing sides with lower high-block and field-tilt scores.

UMAP and Clustering (part 2)

This procedure was also made for the player database.

This database was composed of all the players from the teams of the PPP database, so they go along together. The metrics used differ, but are possession-adjusted in this case as well. I’ve deleted all players that played less than 500 minutes this season and last season too. Figure that will probably increase later in the season.

For players, the UMAP and clustering phase, is in two phases. The first is made for all kind of players, so we find the player position and role on the pitch.

For Amrabat, we look at cluster number 1. It’s composed with defensive midfielders and players that operate in that kind of zones, so we will find players like De Sciglio, Jürrien Timber, and Lucas Hernandez with Kessié, Amrabat, Torreira and many more.

UMAP and Clustering (part 3)

The next phase, is to apply the same procedure to this specific cluster and divide players for specific functions in that role/position. This second phase has been done for every player’s cluster, but this is the result for the one we are interested in:

We used specific metrics for the role in every phase 3 UMAP and clustering. In this case, we are focusing on cluster number 2, which is the cluster I’ve labeled as Regista and in which Amrabat lies.

These are the kind of players that play many medium to long-range passes and operate more in the middle third of the pitch than in the defensive one, so they’re heavily involved in possession.

So, who is it gonna be? The names.

So all of that work, what was it for?

Well, this work - even with his flaws - lets us filter in a fast way players that play in a team with a playstyle that resembles one of the teams we’re working for. After that, i allows us to easily find either similar players based on a profile or functions on the pitch.

It gives us a list of players to deeply scout with ease and speed.

The list

When we apply all of that work and filters we have talked about above, we have this list, which is a further filtered list of only players that are born after 1996, play in teams that are similar to Fiorentina this season, and share Amrabat’s functions on the pitch.

How to go further from here?

The next step is to find the best one. To do this, I’ve decided to use five different metrics:

• True Tackles and True Interceptions possession-adjusted, because obviously we need someone who can regain possession and protect the defensive line.

• Successful Passes %, self explanatory.

• Passes into the final 3rd and xT, because we want players able to progress the ball in valuable positions and help the team be threatening.

So, we take this five metrics and find the median value for our list in each one. After that we create a dataframes for each one, but only with players that are above the median in that specific metric.

At this point we look at how many times players appear in the above median dataframes.

Result

We find that Torreira appears only three times above median, Amrabat four. Four times also for Sadílek, Edson Álvarez, Santiago Sosa, Vitor Carvalho and Alberto Angulo.

Vitor Carvalho deserves a special mention because he appears in the list for both 21/22 and 22/23 seasons, so he would probably be the fourth choice.

But now we go with the top three, the only ones that appeared for all of the five metrics, from third to first.

Federico Navarro

The Argentinian born in 2000 is the youngest of the three and probably the less experienced.

He actually plays in the MLS and this would probably deter many European clubs while going in the market, even more Italian clubs that seem often very fearful on the market.

He probably needs a bold suitor, but deserves a move to prove himself in the so called “bigger leagues”.

Ali Musrati

Ali Musrati, Libyan, 1996.

The most experienced and old of the pack, he currently plays in Braga. He’s Portugal League proven and UEFA competitions experienced. He is probably the safest bet, but the one with less potential, consequently, the one with less development to extract from him and the later resale.

Ramiz Zerrouki

My favourite of the three.

He’s the perfect equilibrium between the previous two, he’s born in 1998, so two years older than Navarro and two younger than Ali. He has two years of experience in Eredivisie and with Carvalho is the only one who appears for both season in the list.

Not only that, he can improve and be sold higher than Ali. He is more proven than Navarro, also, he’s Algerian so potentially he could ask information on the club to Amrabat himself. The last important thing to know about him is that all of his five metrics improved from 21/22 into 22/23 season, he ranked for each one at least forth for the current season.

He would be my preferred choice to replace Amrabat.

Thank you for reading “Who your Football Club should sign?”. This post is public so feel free to share it.