Analytics And Anecdotes – Football

Just so we’re clear at the outset, I am NOT talking about American football- you know, Tom Brady, Aaron Rodgers and the like. I am talking about football which is actually played with the foot. ūüėõ

Analytics and Anecdotes is a four part series whereby I attempt to examine the growing influence of statistics and analytics in some sports -popularised by the Brad Pitt film, Moneyball- while it’s apparent inability to permeate existing structures of assessing players in others. In the first part, we discussed the basic tenets of analytics. In the second, we looked at how baseball has been transformed and is now light-years ahead of any other sport in it’s use of analytics. And finally, in the third, we looked at statistics in basketball and how far it is both embracing and sceptical at the same time to view the truckload of data there.

In the final part of the series, we look at the beautiful game. For a game that is by far the most popular in the world, football remains surprisingly behind other sports in terms of it’s utilisation of statistics as opposed to scouting, despite the many obvious problems associated with the latter. In contrast to the MLB and the NBA, the failure of statistical advancement in football is evident. Why is that the case?¬†

article-2406576-1b4ff39b000005dc-172_964x720
Jamie Carragher and Gary Neville before their days as Sky pundits

FC Midtjylland – A Case Study

FC Midtjylland is a club in the Danish Superleague. In July 2014, Brentford Football Club owner Matthew Benham took an interest in the club. For those of you unfamiliar with Mr. Benham, he made millions by using mathematical models to predict the results of football matches. Under his ownership and an amenable coaching staff, Midtjylland secured their first ever Superleague title in 2015 and missed out on qualifying for the group stages of the UEFA Champions League by losing to Manchester United- not bad for a team of their stature.

Midtjylland and by extension, Brentford remain, however, minorities in the footballing world. In fact, when told by journalists that Brentford was going with mathematical modelling, the response of Micky Quinn, a scout at Wigan Athletic was “They want a head coach and mathematical modelling ‚Ķ Ha, ha, good luck with that.” Despite Brentford achieving their aim of going to the Championship from League One and making the Championship play-offs in 2015/16, many balk at the idea at that statistics could win titles.

Professional football, in comparison to professional basketball has a long and storied history. And for generation after generation, filling up the payroll of the club involved personal identification of players who have a natural ability on the ball. Players like Pele, Diego Maradona, Peter Shilton, Alfredo Di Stefano and Eusebio were identified by hard-nosed scouting. But good old scouting is a hit-and-miss. For every success story in football, there are more than enough number of failures.

Before we delve into why stats have failed to gain as much importance as scouting in football, we’ll look at some of the factors that have influenced the rise of statistics in world football- even if it may not be as phenomenal as baseball

1. A Global Game And The Limitations of Scouting

Even for the biggest clubs in the world, like Real Madrid, there is an ever-present problem with scouting. The simple geographical scale of football and where it’s played in the world is a problem for managers and club owners when it comes to identifying talent. Today’s football is competitive with the drive to find the next Cristiano Ronaldo and the next Lionel Messi more than ever in the game’s history. In these circumstances, clubs want to be as thorough as possible when they are evaluating and looking for talent. However, as I mentioned earlier, the sheer global nature of the game makes scouting for every single player impossible.

This is where companies like OptaPro and Prozone (which recently teamed up with the¬†Football Manager¬†game series- I’ll touch on that later) come in. In an article on The Guardian, Prozone’s¬† business development director Blake Wooster, who counts Real Madrid and Manchester United among his clients, explains: “It’s like when Amazon tells you other books you might like after a purchase. A coach might not have heard of a player in the Polish second division ‚Äď but he might have similar attributes to the guy he’s looking at in League One. We are just increasing the due diligence process.”

For clubs, the use of data and analytics is a way to increase scouting range and increase the global connectivity with other leagues.

2. The Rising Costs Of Purchase

<> on September 2, 2013 in Madrid, Spain.
Gareth Bale moved to Real Madrid from Tottenham Hotspur for a then-record fee of 100.8 million Euros

For any football fan who has kept track of the game for quite some time, the cost of the game has gone up. From ticket prices to the cost of merchandise, everything is on the rise. Blame Arab owners or whatever, but that is the truth. And the purchase of players has followed the trend. Just consider this. Real Madrid bought Portuguese left back Fabio Coentrao for 30 million euros in 2011. Eight years before, in 2003, they got David Beckham for just 5 million more. Clubs are now demanding more and more money to let go of players- especially those who they think will be genuine global superstars in the years to come (just ask Monaco about Anthony Martial or Lille about Eden Hazard). These demands, of course, take a lot of shapes and sizes, from high initial payments to clauses for club appearances and goals, international appearances and goals and money to be paid for winning certain awards or competitions and so on.

Data analysis enables teams to decide if they should go for a particular player when a cheaper one may be available elsewhere. They help decide, on any given day in a football field against a more or less similar calibre of opponents, who will match-up better. Of course, other factors like the home-grown rules present in every league and the foreigner limitations and work permits present in some leagues also help decide transfers but for the most part, teams are looking for the same or better quality of players as someone they saw by themselves, without the same cost.

3. The Unpredictability of Football

Just as any team sport, football is unpredictable. With the amount of variable factors in the game, like the simple fact that there are twenty two different brains who are playing the game at one time and the fact that what happens on one end of the field (like a run behind a defence on the left wing) can influence what happens on another end of the field (like a midfielder seeing this and playing a pass that splits the defence and meets the runner), it is hard to predict the course of events in a football game.

Statistics allows coaches and players to narrow it down. It gives data on say, things like a player’s preferred foot or how a player reacts when he is going one on one on the wing against a defender- does he cut inside or does he go outside and cross? Statistics and analysis can help players try to influence the game by studying their opponents and the latter’s tendencies.

40611829
Lionel Messi and his Barcelona teammates under Pep Guardiola exemplified the tiki-taka, a system based on positional inter-dependence and possession

Football has a level of positional inter-dependence that no other sport has. For a system to be effective, football needs all eleven players on the field to buy into that system and execute it. How a team performs over a course of ten, twenty or thirty games is indicative of how much cohesion a team has.  As a consequence of players depending on each other to perform and to make the system work, individuality is practically non-existent. The only places where some degree of individuality may be found is when

  1. There is a one-on-one between the winger and a full back on either flank with no support for either player
  2. On counter-attacks when there are very few players involved and large swathes of space to operate in- allowing for more space in which players can bring out and exploit weaknesses. For example, a defender can force an attacker to dribble with his weaker foot by “showing” him onto that foot. This can slow down a counter and it can lead to unforced errors.
  3.  One-on-one situations- This includes both one-on-one situations between a keeper and striker and penalty kicks.
  4. Set pieces- More and more coaches, especially on the international football scene are looking at set pieces- both on defence and on offence- as areas which can be manipulated into scoring situations or into situations for a subsequent counter-attack as the need may be.

While all these factors have contributed to the increased use of statistics and analytics, including by big clubs like Chelsea, Liverpool, Real Madrid and Barcelona, there is a noticeable lack of acceptance among the people involved within football. This can be attributed to a variety of reasons.

One of the more important reasons is that despite whatever stats are available, the simple number of confounding factors over a ninety minute game are too many to allow for some kind of numerical prediction of a match. For example, each player’s performance on the field is affected by his teammate’s performance, by how superior the opponent is, the player’s fatigue level, skill level, the opponent’s skill and fatigue levels and so on. While all these factors also exist on a basketball court, the lack of individuality in football- something which, as we discussed in the last segment, basketball still possesses- and the huge amount of inter-dependency ingrained in the game makes any attempt at narrowing a game to a single predicted score that, more often than not, correlates with the final score is futile.

Another reason is that unlike baseball and even more than basketball, football is a game of systems. And it is imperative that any statistic must be read within the context of a system. For example, in our introduction to this series, we spoke of Leicester City and Barcelona and possession as a statistic. Both Leicester City and Barcelona have become champions of their national league in recent years (both in the 2015-16 season most recently). While Barcelona averaged 62.9% possession, their English counterparts Leicester averaged only 44.8% in the same category- this was a record 18th even within the Premier League, only ahead of Sunderland and West Bromwich. Possession is thus a strange statistic and definitely can, at the same time, be used and not be used to correlate with success. The difference, of course, is that Leicester’s¬†system¬†was based on them attacking on the break and getting the ball moving forward quickly, using the speed of Riyad Mahrez and Jamie Vardy to unsettle defences while Barcelona probed the opposition for a weak spot before striking. Speaking further of systems, the use of statistics to evolve and tweak systems has also been met with mixed results. I am not going to go into the details of it but suffice to say that the more you try to play to your strengths, the more the opposition is going to try and stop you – with, like I said, mixed results.

3772
Eden Hazard celebrates his goal against Manchester City on December 3, 2016.

Another major flaw is the interpretation of statistics. I have a very good example for this actually. During this season’s game between Chelsea and Manchester City, an acquaintance of mine and I differed on what was meant by a “chance”. His argument was that Chelsea had four shots on target, three of which went in- meaning they were taking their “chances”. My argument was that by chance, I mean any situation where taking a shot on goal was the best possible option and that in that respect, Chelsea ¬†and City were both not that great. My definition, I must admit, is based on¬†The Football Manager¬†games where you have a statistical analysis page showing clear cut chances and half-chances. In my opinion, a clear cut chance is where shooting is the best option- a 1v1 for example- and a half-chance is where a shot would have been a good option but there may have been equally good options in terms of a pass. By that respect, I felt Chelsea-especially in the first half of the game- had more chances (both half and clear-cut) than what a simple shots taken in the first half showed (It showed zero by the way). Most notable among them, of course, was when Eden Hazard cut past the keeper and he decided to pass the ball into the box than try a shot at an open goal from a narrow angle but an angle from which scoring was far from impossible. City, of course, dealt with the pass.

If you ask me, a simple shots-to-shots on target ratio in isolation does not do a good job of explaining things to me about a team and it’s efficiency and decision making. For example, A 75% ratio can also be got by three shots from beyond thirty yards if all three shots directly reached the keeper and a fourth shot went wide. That way, you might have a good ratio but you don’t have good decision making. And yet, like the person I was having an argument with, people tend to focus on just simple statistics which hardly do justice to a team. This is true not only of spectators like him, but also many people within the football establishment. This leads to misinterpretation of statistics. Moreover, people are suspicious and dismissive of what they don’t understand. It’s human nature. And the notion that a sport can be mathematically explained- never mind that people only try to explain a part of it- is something that people find hard to come to terms with.

It is, however, making inroads- especially in match preparation. Simon Mignolet, Liverpool’s goalkeeper, had a dream debut when he saved a last minute penalty from Stoke City’s Jonathan Walters. While he puts it down mostly to luck, he also says he studied Walters’ penalties and worked with the Liverpool goalkeeping staff. Mignolet is not a case in isolation. But his is the best example I can take here because of now, he has the best record for any keeper in Premier League history as far as penalty shootout saves go (6 out of 12 shots saved as of the time of writing).

Moreover, as Brentford and Midtjylland have illustrated, data analysis allows them to play Moneyball football. Is Brentford going to win the Premier League title within the next five years? Probably not. But Brentford is definitely punching above their weight in the Championship and should they reach the Premier League, the trend is likely to continue.

However, despite this, scouting retains it’s primary place in the football world. One major reason, as OptaPro’s Simon Farrant said in the same interview to the Guardian, “We can’t tell you how a player behaves off the pitch.”

screen-shot-2016-11-01-at-15-21-30
A screenshot from Football Manager 17 showing an interaction between two coaches via the media.

However, even this is increasingly coming under threat from analytics. Although it still has not surpassed fairly rudimentary stages,¬†Football Manager, the world’s most popular simulation game series about football management uses things like player characters – whether they are loyal, ambitious, volatile with the media, professional and so on- to judge whether he will be a success with a particular team. And this goes not just for players, but coaches too. FM uses the cumulative system of mental and personality traits to determine whether a person will be a success or a failure at a club under a certain coach with a particular set of teammates. All this, in addition to his on-field ability.

Another reason scouting remains a mainstay in judging talent is actually one of the things I mentioned before. Talent and players, must always be judged in the environment it is displayed in. To put it simply, one needs to look at context. And while numbers may give you a certain idea of how good a player is, contextual visualisation is important to judge if a player is being made to look better than he is or if he is genuinely talented. And this can only be done by watching him and his teammates play and not by simple numbers.

The final factor that I would touch on is the fact that statistics in football are not developing at a huge rate. Baseball has almost completed it’s development and basketball statisticians are trying to break down the game on a daily basis but football statisticians still focus on trying to make meaning out of existing statistics instead of trying to derive new ones that make sense and can explain the myriad of intricacy that go into a football game a little better. Part of the problem is¬†that most of the people who are good statisticians and whom teams can really make use of, are employed by the football betting industry. And secondly, for anybody going to make mathematical sense out of football, it is a trying task to find connecting dots among the myriad of intricacies that adorn a football game.

The use of statistics in football is in a stage of infancy- a stage where it is still finding it’s feet. While success stories like Brentford and Midtjylland exist, there are also failures like Liverpool (though it isn’t strictly moneyball football, it qualifies on a loose definition of the term). Moreover, the complications that exist with the sport of football in terms of the number of variables and the number of human factors – all of which are controlled by different people- involved, are far too many to neutralise through a simple equation or an algorithm. However, there is no doubting that statistics in football, over the next few years – even decades – will gain far more importance than simply being used as a simple shortcut to a talent list shown on a computer screen.

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s