Some football Analytics!

It's really hard to ignore the football fever going on around you. I don't claim to be a big football fan - I know a few names here and there - but I usually keep my mouth shut with far more dedicated fans around me. 

What I like about the game is the level of coordination that it requires - probably the highest in any game. If played tactically, it can engage 22 people at the same time. Sometimes it even feels like chess in motion. One thing struck me about this FIFA was the alarming exits of great teams like Spain, Italy and Spain. The top teams like Brazil has not done well particularly - not played like a champion. 

So I started wondering whether co-ordination can really matter. Say Team A has players, of whom a majority of players play or have played for the same club in the recent past compared to say Team B, with a larger number of players belonging to a different group - would Team A have a higher chance of winning? Why do I say so?

1. The more time a player spends with another player - the more they are able to co-ordinate their moves. This is quite evident. It is like knowing the moves of your dance partner so that you are able to time and judge your moves in a better way. For example - A center-forward player would become accustomed to the moves of players on his left and right. The way they co-ordinate - pass direction, pass ball speed, pass timing have a large impact on the probability of scoring. 

Also, players trained under the same manager will be adept at using the same tactics. This may range from anything such as – playing attackingly/defensively, nature of passes (short/long) etc.). If a team as more players who have played in the same club under the same manager, they would have collectively mastered a team strategy that they can find difficult to execute with players from different clubs.

2. In continuation with the above hypothesis, it is assumed that, given the skill level of a player, he will take more time to adjust to a new "partner" (the partners can be multiple - considering the players in his vicinity), the less he has trained with the new guys (and the more he is accustomed to the older partner). The 1 & 2 points can be looked together. Another extension of this, more practical and based on how usually players are distributed across clubs, is to consider that a good club (ManU, Chelsea) would have 3-4 star players in them. In a national team, they are less likely to find such players in their national team in their vicinity. 

I say this because, by a star layer I mean, one among the Top 15 players spread across 3 or 4 clubs - meaning 3 or 4 players in each club. However, more than 10 nations compete in FIFA and we end up observing that each nation usually has one star player - Neymar (Brz), Messi (Arg) etc. In light of this argument, the above point makes sense. 

3. Players will develop more rapport over time. This would usually mean that they would gel better as a team and the ego clashes would be lesser.

What I intend to measure

1. The first task would be to quantify the level of co-ordination that exits in the national team - measured using the following data. (Although, many other parameters exits - successful pass, passes converted to goals etc. - but I intend to identify one of the possible causes of this)

1. Current team composition of a team.

2. The following player data -

• Position the player plays in the national team.
• The club(s) the player has played for in the last 4 years (interval between FIFA) and the position(s) he has played in. 
• The number of matches the player has played for those clubs.

The net level of co-ordination of the team would be measured as follows -

Where, i and j are indexes for players running from 1 to 11 each and i ≠ j

K_i,j is 1 if both the players play for the same country.

As pointed out earlier, the effect of playing for the same club has two impacts. The first one is better passes and knowledge of the players surrounding him. To measure this, we need to use the position the player plays in when he represents the national team as well as when he plays for the club.

For simplicity, let us mark the positions a particular player can play in from 1 to 10 (ignoring the goal keeper for simplicity, for now. I’ll discuss that later as this has a completely new dimension to it.)

Let p_i (Cl) and p_i (Nt) represent the position the player plays or played for in a club and the national team respectively.

We should add one more assumption here – A player is very less likely to co-ordinate directly with a player that is more than 3 positions away from him . Thus the following formula is established –

For | p_i (Cl) - p_i (Nt)| < 3, (if the players are separated by less than 3 positions)

Else, 0

The formula would work as follows – Say we have a 4 year data on all the club matches that happened, the players who played in those matches, the positions they played in and their nationality.

For each pair of players in a particular match it would pick up each side (a club) and proceed as follows.

1. Choose a pair of two players.

2. Check the positions they play in and check if the separation is < 3. If yes, proceed, else exit.

3. Check if the players play for the same country. If yes, proceed, else exit.

4. Add 1 to the score of the team of the player (or players).

5. Repeat.

Call the net score of the team as “Co-ordination score”.

The second level is easier to calculate and focuses on how easily the players will be able to gel. Instead of measuring the matches played, it makes sense to simply look at the time spent together at a particular club. Hence, the same formula applies, with the matches replaced by time spent. Call the net score of the team as “Bonding level”.

The overall score of the team is – w*(Co-ordination score) + (1-w)*(Bonding level)

This is half the story. I started by positing that teams which have more co-ordination and bonding, which comes from spending more time and playing together at the same club, makes them perform better. I said that these teams, I believe, would have a higher chance of winning. I would make it more specific now.

It is understood that this can be, if at all, be one of the many drivers of success of a team. We have observed that teams have won with just one star player seeing them through. Thus, trying to relate the winning possibility of a team with the co-ordination levels would be a bit far-fetched. It would make more sense to use a lower level parameter such as total number of passes or passes that resulted in a goal.

These variables will directly measuring the level of co-ordination that the team has on the field. It is interesting to note that often the level of gelling among the players can be a huge driver of win. Thus, although a higher co-ordination score might translate into more accurate passes on the field (to be tested), Bonding level might actually determine the overall outcome of the game (better communication on field and in the locker room).

The hypothesis to be tested

Null hypothesis 1 – A higher overall team score increases the chances of a national team winning in FIFA.

Null hypothesis 2 – A higher coordination increases the number of successful passes at the team level.

I’ll try to test this on a small dataset. Doing this for several years would require some coding.

The goalkeeper issue

I earlier ignored the mighty goalkeeper. This requires special treatment. Consider a FIFA match between Brazil and Argentina. Say the goalkeeper of Brazil played for Club A, where 5 of the players from the Argentina national team also played with him.  The problem is problem reversed – playing together is bad for both the national teams.

The players from Argentina know whether how the goalkeeper moves during penalties (they are able to anticipate him better as they regularly train with them), they know their weak spots and may target him specifically.

Similarly, the goalkeeper knows whether during penalty the Argentina players are more likely to hit towards the right or left (this can be eliminated if the players randomize – an interesting example of a 2X3 simultaneous game), may know his other weaknesses that might be helpful when he plays for the national team.