Scouting Portugal’s 2nd Division for a Matić Replacement

Nemanja Matić has been a top defensive midfielder shielding the defence and building up play over the past decade. He has served Manchester United well during his tenure at the club, but they will soon need to find a suitable successor for him as he is now 32 years old and past his best years. Even Chelsea had got in Kanté and Jorginho to replace Matic a year prior and after his departure to United. He has performed admirably for them over the past three seasons as a defensive midfielder shielding the back four of their defence. However, United have a lack of players like Matic who can perform his role effectively. Also, United only have two other first-team players who are defensive midfielders in Fred & McTominay, neither of whom are exceptional nor of a similar mould to Matic. Given that Matic has recently renewed his contract with Manchester United for three years till June 2023, I have assumed that they will keep him around for that time period. During this period, United could get in a player of a similar mould to Matic, who can help groom this player, and who will take over from him.

Thus, I decided to use data to scout for and find a replacement for Matic using Wyscout. I have selected the Portuguese 2^nd division to carry out this exercise as a case in point to show how data that is available for lower-level leagues, which have lesser coverage can be utilised for scouting.

First, I got the raw data of Matic, Big 5 European League and Portuguese 2^nd division central midfielders who have played at least 500 minutes in the 2019-20 season. Then, I computed my own derived and relevant metrics from those, for which I also used Principal Component Analysis.

I used the metrics to find out what type of player Matic is by checking what type of role he falls under. For this, I have defined multiple types of roles for central midfielders (as well as multiple roles for each other position). I came to these by identifying the different types of roles performed by different types of top players playing in the same position and analysed the top ideal players for each type of role. I identified the important metrics out of the available metrics from Wyscout and those derived by me, for each role under each position. I also tested this method and my identified metrics using radar charts and k-means clustering to see whether players having the same role have similarly shaped charts and fall under the same cluster.

I used data to filter and cluster players in my dataset into different roles, to see what type of role Matic comes under. Matic fell under the cluster of the Anchor role, which matches my intuition and is verified by my eye test. (I define an anchor as: A protecting midfielder playing in a deep position close to the centre backs, who is dominant aerially, strong at tackling and/or interceptions.)

After establishing that Matic is an anchor, though one with good passing ability, I moved over to the dataset of Portuguese 2^nd division central midfielders. Using the relevant metrics for anchors I have defined, I filtered the data to find the players fulfilling the basic required criteria for an anchor such as game reading and defensive/aerial duels. I further filtered the data for players aged below 30, given that I am looking for a young replacement for an ageing player and the transfer strategy being followed by Manchester United. After going about this process, I came down to a shortlist of 11 players out of the initial 76.

I then looked at the profile and all the relevant metrics of these 11 players, to see which of them are better players and more like Matic. After comparing Matic’s relevant stats from the past three seasons through a radar chart, I also compared each of the shortlisted players to Matic using radar charts. In the radar charts, I have considered only successful attempts normalized per 90 minutes and taken the minimum and maximum value for each metric as the 5^th and 95^th percentile among all the relevant central midfielders. In case the value of any metric for a particular player is an outlier, I have rescaled the maximum value for that player to fit it within the radar chart.

Overall for his role and position, especially defensively, Matic had his best performance for United in the Premier League in the 2017-18 season. I also decided to identify three pros and three cons of his.

Pros	Cons
Game Reading	Sliding Tackles
Running with the ball	Shooting
Passing	Age

Below I have compared Matic’s statistics from the 2017-18 Premier League season against each of the 11 shortlisted players from the Portuguese 2^nd division:

I then analysed Matic and 11 of these shortlisted players using X Y Bubble charts to gauge their different abilities such as aerial prowess, game reading, discipline, defending ability, passing style.

Note: In all the XY Charts the red dashed vertical and horizontal lines demarcate the 95^th percentile (among all central midfielders as mentioned earlier) for the metrics represented on the X and Y axis. The only exception is the discipline chart in which the lines demarcate the 75^th percentile. The 11 shortlisted player names are coloured green, Matic’s name is coloured red and all other players above the percentile lines are coloured white. Each XY graph measures a certain ability and in each, any player(s) fairing poorly will be flagged. If a player is flagged two or more times then he will be filtered out from the shortlist.

From the above chart, I can see that all my 11 shortlisted players bar Savio engage in more aerial duels per 90 minutes than Matic. All of them win more than 40% of their aerial duels with Savio, Alhassan, and Luis Silva falling above the 95^th percentile and winning more than 67.9% of their aerial duels. Only Rafael Brito & Sergio Marakis have a success rate below 44%, but given that Brito only turned 18 years of age this year, he is likely to develop physically and improve his success rate. I can deduce that all the shortlisted players, except Marakis, are capable in the air and have a decent to strong aerial prowess. Thus, I will flag Marakis.

I can see that Matic is an adept reader of the game as he has a high number of interceptions and shots blocked. Savio and Ramires excel at game reading, with Ramires having the highest interceptions and shots blocked. Luis Silva is the worst on my shortlist, fairing really poorly in this essential measure, so he is flagged. The other players on my shortlist are capable game readers, though Diogo Mendes is far from the best. 

For this exercise, I have taken Matic’s progressive passing numbers of the 2019-20 season. I can see that Matic completes more progressive runs than all my shortlisted players and only one, Diogo Mendes, completes more progressive passes per 90 minutes. The three players who can progress the ball are Savio, Rafael Brito and Diogo Mendes. Sergio Marakis and Filipe Cardoso fair poorly, with the latter being flagged and the former being removed from my shortlist.

Matic attempts more progressive passes than all my shortlisted players but doesn’t attempt the most as a proportion of total passes, attempting fewer than 15 progressives passes per 100 total passes, which is indicated by the size of the bubble. This shows that Matic tends to have a lot more possession of the ball as compared to my shortlisted players. Three of my shortlisted 11 players aren’t good progressive passers, namely Marakis, Cardoso and Pedro Ferreira. Cardoso is the 2^nd to be removed from my shortlist with Pedro Ferreira being flagged. 

I can see that Matic is extremely well-disciplined committing a low number of fouls and receiving a low number of yellow cards. None of my shortlisted players is close to being as well disciplined as Matic, but I will use this chart to see which players lack discipline and fall above the 75^th percentile. Savio is the rashest and most undisciplined, with Tembeng, Ferreira, Brito, Cardoso and Marakis also being poorly disciplined. Ferreira will be removed from my shortlist, with Tembeng and Brito flagged.

From the above chart measuring defending ability, I can see that all my shortlisted players, other than Savio, Marakis and Cardoso have good to great defending ability. Savio is the 4^th player to be removed from the shortlist.

Matic has great stamina as he completes a higher number of progressive runs than my shortlisted players and plays the 2^nd highest number of minutes per match. Savio strikingly low minutes per match indicate that he rarely completes full matches and is either subbed on or off. Ramires and Alhassan will be flagged as they complete a very low number of progressive runs coupled with playing fewer minutes per match than most of the shortlisted players.

This is a chart that measures the passing style of players, which will indicate whether they attempt long passes often or tend to keep it short & simple. I can see that Matic doesn’t attempt a high proportion of long passes. All my shortlisted players have a fairly similar passing style to Matic.

From this chart, it is clear that Matic is a brilliant short passer of the ball as he attempts an extremely with a brilliant success rate. The shortlisted players fairing poorly on this are Marakis, Ferreira, Tembeng and Savio. As Tembeng is flagged for the second time, he will be removed from the shortlist.

Matic attempts more long passes than any of my shortlisted players with a high success rate. Ramires, Marakis, Tembeng and Pedro Ferreira aren’t good at passing the ball long as indicated by the combination of the number of long passes they attempt and their success rate. Ramires will be removed from the shortlist.

The above chart representing goal contribution is not an essential one, but a nice to have one. As expected, it shows that Matic, as well as most of the shortlisted players, have very low goal involvement in terms of goals and assists. Only Ramires and Savio have a meaningful goal contribution.

After looking at all of these X Y graphs, the radar charts and the statistics of each player, I further reduced the shortlist to 5 players viz. Rafael Brito, Ndiaye, Diogo Mendes, Alhassan and Luis Silva. The other 6 are either not stylistically similar or good enough defensively, having been eliminated from the shortlist using the XY charts.

I once again looked at each of the 5 in more detail, listing down their pros, cons and analysing who would be the top replacement(s). 

Brito (Benfica B)

Pros	Cons
Interceptions	Running with the ball
Defensive ground duels	Creative/risky passing
Passing ability	Discipline

Ndiaye (Porto B)

Pros	Cons
Aerial duels	Ball progression
Game reading	Creative/risky passing
Discipline	Running with the ball

Mendes (Benfica B)

Pros	Cons
Aerial duels	Long passing
Running with the ball	Creative/risky passing
Discipline	Game reading

Alhassan (Nacional)

Pros	Cons
Aerial duels	Ball progression
Game reading	Creative/risky passing
Discipline	Running with the ball

Luis Silva (Leixões, now Chaves)

Pros	Cons
Defensive ground duels	Game reading
Slide tackling	Creative/risky passing
Aerial duels	Running with the ball

I also carried out a clustering exercise to group each of the Portuguese 2^nd division central midfielders and Matic into clusters using K means clustering, which groups similar players together. I checked whether the players fall under the same cluster as Matic and indeed are similar to Matic.

I saw that all 5 of my final shortlisted players, Rafael Brito, Ndiaye, Diogo Mendes, Alhassan and Luis Silva are all in the same cluster as Matic.

Finally, I conclude by recommending Rafael Brito as the successor to Matic from this league, as he is the best like-for-like replacement and most similar stylistically. He is already performing at a good level having turned 18 years old only earlier this year, has high potential and a lot of scope to improve his already good performances. Interestingly, a Google search on his name also shows that he has already been linked with moves to Liverpool, Barcelona and Real Madrid!

P. S. Check out my football merch business at midfieldmerchandise.com