FM21: Les Bleus — Exploring the AI behind national team selection.
The Problem
It dawned upon me the other day that to date, despite winning the UEFA Europa League in my first season in charge of OGC Nice, running PSG to the wire in my second season and progressing to the latter stages of the UEFA Champions League, Les Aiglons are yet to have a single representative called up to Les Bleus.
This piece has the aim of establishing how artificial intelligence is used within Football Manager to select players for their national sides.
Hopefully providing me with a better understanding of how Didier Deschamps is making his squad selections and improving future chances of getting players selected for the national team.
This is my attempt to force Deschamps hand to include a player from OGC Nice in Les Bleus.
Les Bleus Starting XI
Above is the current starting line-up for Les Bles, the eleven likely to feature during the World Cup in Qatar.
Analysing the Data
Squad Profile
The squad profile will be my starting point in this dive into selection, I have opted to explore minutes played first and foremost, as minutes played is the only metric which I believe the game flags you with a warning in relation to selection.
I am sure many of you have received a message stating that one of your players may be at risk of not being selected for their national team due to not playing significant minutes for the club.
Above I have created a squad profile, if you are keen to make one of these yourself I have created a how-to which can be accessed here. This shows the minutes played for the current French national team at club level and includes French players from OGC Nice who have featured in over 1,000 minutes across the 2021/22 campaign.
Analysing the data visualisation, there is certainly a positive correlation between minutes played and selection for your national side. As you can see the majority of players within the Les Bleus have played close to 3,000 minutes for their club sides, with the exception of a few players (Dembele and Tolisso both suffered moderate injuries over 2021/22) and Alphonse Areola, the second choice goalkeeper who played zero minutes!
This leads me to believe that out of the 10 players who have played over 1,000 minutes for the club only the three which have secured over the average minutes will be in scope for further analysis.
The candidates are;
- Ludovic Blas
- Stanley Nsoki
- Khephren Thuram
Metrics
Using the above information as a baseline I will now look into the per 90 metrics which sit behind these individuals to compare how their performance on the pitch compares to their national counterparts.
This will hopefully provide an indicator as to whether performance is considered in the selection process and if so how much of a weighting it holds.
Centre Midfield — Passes per 90 vs Key Passes per 90
Starting with the Midfield, an area in which I have two potential candidates. Firstly, I have opted to compare passes completed per 90 with key passes per 90. These two metrics I feel are essential for a midfielder, particularly in a team who look to dominate possession like France.
Analysing the data you will see that both Aouar and Tolisso are both performing above average in relation to passes completed per 90, with Aouar leading the way in terms of key passes per 90.
Yes I know that style of play at club level will influence these figures, however taking from it on face value Aouar is the creative of the two midfilders in France’s top down formation and Tolisso operates in the box to box role.
This makes Aouar, Blas’s direct rival for the shirt, who operates for OGC Nice in the role of the advanced playmaker.
However we don’t want to jump the gun, Blas isn’t even being selected which makes him need to be better than Rabiot who is an impact player for the team, he is doing this against both metrics.
Thuram on the other hand is leading the line in terms of passes completed per 90 but as a deep lying midfielder we will need to explore the defensive metrics to understand why he isn’t being selected.
This will also give us a better understanding on the reason as to why Rabiot might be getting the call instead of Blas.
Centre Midfield — Tackle Rate vs Interceptions per 90
I have also opted to compare tackle rate and interceptions per 90, these metrics are to focus on the impact of the midfielders without the ball and will favour the more defensive player role such as the defensive midfielder.
As you can see from the above, Kante is an extreme outlier in terms of his interceptions per 90 whilst Camavinga has the best tackle rate in the sample.
The French midfield trio all have one thing in common, they all are performing above average in terms of interceptions per 90 an area of the game which both Thuram and Blas are underperforming.
Rabiot is also performing above average against this metric, could this be the decisive factor in relation to him being included?
Central Defender — Headers won per 90 vs Tackle Rate
Taking a break from the midfielders we move focus towards the performances of the central defenders to see if we can shed any light on the selection of the defensive unit.
This time my focus is on headers won per 90 and tackle rate.
Our players do not compare well across these metrics, it is clear to see that Deschamps has a preference to select players who have an aerial presence with all his central defenders performing above average against this metric.
If our best candidate Stanley Nsoki is to have any chance of selection maybe I should move him from his role as a ball playing defender! As taking just the tackle rate he is on level par with Varane and leads the line in terms of performance against this metric.
Attributes
My final area for analysis is the in-game attributes, this will give us an insight to see if players are picked purely on current ability. In order to do this I have picked a range of key technical, mental and physical attributes needed for each role and calculated an average score for each player.
Midfield attributes
The average score for central midfielders below is based on individual scores in the following attributes:
Technical: Passing, First Touch, Technique, Teamwork
Mental: Decisions, Off the Ball, Composure, Vision
Physical: Stamina
Looking at the attributes it is clear to see that 15 or above is a exceptional score which will lead you to having a top class player.
Thuram is bottom of the pile in terms of his average in relation to these attributes, however Blas is outperforming both Tolisso and Kante.
With Rabiot only slightly better in terms of his attribute rating it is a positive sign for Blas who must be within touching distance of a call up to Les Bleus.
Defender attributes
The average score for central defenders below is based on individual scores in the following attributes:
Technical: Heading, Marking, Tackling
Mental: Positioning
Physical: Jump Reach, Strength
As expected with the defensive unit our players despite performing well within Ligue 1 aren’t quite there in terms of the spread of attributes in comparison to their national team counterparts.
In fact he is averaging 1.5 points less than the top three centre backs, considering this is across six attributes a greater emphasis will be put on his development over the coming year to attempt to bridge the gap.
The Conclusion
It would seem that attribute scoring and minutes played has the biggest influence on selection for the national team. Given that the two have a positive relationship and the likelihood of more minutes will see an improvement in attributes is music to managers ears.
Play the man and he will improve his attributes!
However by drilling down on the metrics you will gain a better insight into what roles national players operate in and how adjusting player roles and instructions may benefit them in terms of securing the national shirt.
I hope you enjoyed this very bespoke piece, it is strange how these blog ideas just jump into mind! RT greatly appreciated on Twitter and applauds on here!
Big love,
Magnus
