# Understanding Zizou’s Success: How Squad Rotation Can Be studied

## Vignesh Velu – @noobiefootball

I developed a model to study squad rotation. I looked into every team’s squad rotation in La Liga.

The playing time of players is always tricky. Too much players will end up having injuries or exhausted for the final stages. Too little it affects the players’ development or their confidence. It is always essential to find the right amount of minutes. They are not Supermen (apart from a few). It is a far more critical thing if you are in multiple competitions. A player becoming injured without a clear replacement will make it worse for teams.

Real Madrid defended the Champions League successfully for the first time in the competition’s history. They also won the cup double for the first time since 1958. One of the key parts in his success was his squad rotation. He managed to give an equal amount of minutes to almost all his players. This article was originally intended for just Real Madrid’s team. But it went onto become a model that could help us understand rotation.

To measure squad rotation, we typically see the average minutes per player. But it is distorted for many reasons. In fact, I was surprised to find it was almost totally opposite. The first problem I had to deal was number of players. Every team is allowed to play any number of players in league competitions, but not the Champions League. The teams generally give youth team players some minutes. The first decision I took was to remove all the players who have played less than 200 minutes. Unfortunately, it didn’t do much as you can see clearly from the table below.

The table still has problems that skew the result. I ended up with three problems to solve.

1. If a player played 90 minutes against a top 6 team and a bottom team, it still accounts for the same minutes. The minutes against the top 6 teams are always better than minutes against the bottom teams. This method doesn’t take that into account. 90 minutes against Real Madrid and Granada are essentially the same in this.
2. Each team still has varying number of players, from 19 to 28.
3. There is also a significant problem with averages. Let us take two scenarios. Each scenario has two players. In scenario 1, player A gets 90 minutes and player B gets 0 minutes. In scenario B, player A gets 60 minutes and player B gets 30 minutes. The average of both these scenarios is still the same (45 minutes). It does not express how much the minutes were distributed among players.

How do you solve these problems? I decided to divide it into two parts. Part 1 solves the 1st problem. The other 2 problems are solved by part 2.

## Part 1: Quality Minutes

This part differentiates the minutes by the quality of their opponents. Essentially it is a quality factor that you multiply to the minutes each player got against that opposition team. If a team played against Real Madrid, it will be the highest. If a team played against Granada, it will be the lowest. One thought I had was xG (expected goals) and xGA (expected goals against). The problem with this approach is that the difference will very low between consecutive teams and sometimes change the position of the table. It will be like a curve if plotted but I needed a downward slope to differentiate teams especially consecutive ones. It may be perfect for other things. Both these values are perfect only in an ideal world. Teams tend to overperform or underperform their xG or xGA. Manchester City had the second best xGA and highest xG. But people watching know that their defence was bad and the opposition team don’t need a big xG to score goals. Real Madrid overperformed their xG and xGA to win the league. It doesn’t count in some things that matter. I am not against xG or xGA. It is still the far better thing to compare performance, but just not rotation.

Another thought was based on the ranking of the teams. Each team has to face 19 different teams. I spread these 19 teams from 0.95 to 0.05 by keeping a 0.05 difference between each team. I multiplied these factors to the number of minutes each player got against different opposition. For a team like Valencia, if a player played 90 minutes against Real Madrid, it will be accounted as (90*0.95) 85.5 minutes. If the player played 90 minutes against Granada, it will be accounted as (90*0.05) 4.5 minutes. Important to note that it will vary for each team. For Real Madrid, Barcelona will the highest quality opponent they could have played against and hence given 0.95 for every minute played against them. I then summed up the minutes each player played. I took the percentage of this total quality minutes to the actual minutes each player got. A player who got 50% essentially means playing the total minute against a 10th placed team. I am not saying it represents the exact quality of the opposition. I’d appreciate it if someone could come up with some better way to show the difference between top teams and lower teams. I also thought about points won which also has the same problem as xG. You can also apply this to Champions League by using European Club Index (ELO Ratings).

Quality minutes = minutes played * quality factor of the opposition;

Quality percentage = total quality minutes / total minutes * 100;

There are some surprising observations when I looked into each team which we will see below.

Despite all the rotations, Ronaldo still managed to get the most minutes. Nacho, despite playing more has fewer quality minutes suggesting how much he was used against lower order teams. Daniel Carvajal has the highest quality percentage (59%). Danilo being his replacement has seen just 41% quality percentage. This clearly suggests the former was used more versus top teams. Kovacic played almost same minutes as Casemiro and more than Isco. Players who were under or near 40% are the ones who have been sold. Morata, Danilo and James rightly went to other teams as they have been looked over for playing against most top teams.

### Barcelona

There are 5 players who have more minutes than Real Madrid’s highest. Iniesta has been used less but against good opposition as he has the highest quality percentage. Andre Gomes has got more minutes than Iniesta but against weaker opposition. Denis Suarez has been given way too less time to prove himself. It looks like Rafinha should have been sold, as he has been played the least against quality opposition.

Atletico Madrid had one of the worst rotation policies in La Liga (which we will see later). There are 8 players who played 2500 minutes suggesting Simeone used almost the same squad week in and out. He used only 19 players. For a club which competed on three fronts, Atleti should have done more rotation. Gabi (33 years old), Filipe Luis (32) and Diego Godin (30) should have been rested more, especially the first two. Barcelona’s board has been criticized a lot for leaving a bad squad but Atletico Madrid’s board deserves some too, for their lack of quality options. Their transfer ban and stadium move this season, obviously, won’t help. Unless they act, they will lose their status as one of the top 6 teams in Europe. They need to add significant quality either through transfers or through their academy. Saul is sort of underrated. He has become one of the most important players at just an age of 22. Especially considering his health (kidney problems) issues, his fitness is great. What a career he will have if he just keeps going!

That concludes the observation for the top 3 teams. Now on to the two remaining problems.

## Part 2: Standard Deviation

To combat the disadvantages of using an average, I used standard deviation. Standard deviation allows you to understand how the minutes are distributed from the centre. It is effectively like an upper and lower bound. It also neglects the one or two players who play a lot or play very less thereby solving the second problem. We will take the same scenario that was discussed in the third problem. For the first scenario (player A gets 90 minutes and player B gets 0 minutes), the standard deviation will be 45. For the second scenario (player A gets 60 minutes and player B gets 30 minutes), the standard deviation will be 15. The average of both these scenarios were 45 minutes but standard deviation showed which was better. Less the standard deviation, more aggressive the rotation was and vice versa.

Real Madrid outdoes both their rival and lead the table. Zidane has certainly rotated his team better than the others. Barcelona is in the middle, and Atletico Madrid is 18th. Atletico Madrid and Barcelona topped the average minute’s table which didn’t show the true state of the club. To my surprise, I found that more the average minutes, less likely they were rotated. Apart from Real Madrid, all the top 5 teams in the first table were in the lower half in the second table.

Even in this table, counting only games in which the top 7 teams last season played each other, it’s obvious that Real was the most-rotated side.

Zidane has definitely mastered the art of rotation. His tactical nous may look limited to some, but he is doing many other things that are essential to a club’s success. Anyone can point out the abundance of talent he had, but he used his talented side effectively. He is surely going to become a legendary coach. He is a humble man and he deserves tons of credit. And to think what more he could achieve…

I am neither a math guy nor a stat guy. That’s why I kept it simple. I would like to thank Ashwin Raman for helping me with this article.