The scowl onRonaldo's face made international headlines when he was pulled from a match with 18 minutes left to play. He is not the only one who feels that way. Few football players agree with a manager substituting them in favor of a new player.
There will be a more evidence-based way for players to argue for time on the pitch during the World Cup. Each player will receive a detailed breakdown of their performance after the final whistle. The strikers will be able to show how many times they ran. The opposing team will have data on how defenders hassled them.
The beautiful game is invaded by numbers. Data analysis helps to steer everything from player transfers and the intensity of training to targeting opponents and suggesting the best direction to kick the ball.
Footballers face data scrutiny more often than astronauts. The number of shots taken with each foot can now be counted with the help of the vests and straps. There are cameras at multiple angles that capture a lot of information. Most elite football teams now employ data analysts, including mathematicians, data scientists and physicists plucked from top companies and labs such as computing giant Microsoft.
In return, insights from analysts are altering how the game is played: strikers shoot less frequently from a distance, wingers pass to a teammate rather than cross the ball and coaches try to win possession high up the pitch.
Daniel Memmert is a sports scientist at the German Sport University Cologne. It has changed the way teams look at opponents and how they look for talent.
Data changing sports comes from a different game. Billy Beane relied on player statistics to deliver a winning baseball team on a shoestring budget in 2002. Detailed data about a player's performance, such as how often a batter makes it to base, was used to recruit players.
Women are more likely to suffer sports concussions.
Beane had an advantage over the others. Memmert says that football is more complicated than baseball. Baseball is a stop–start game in which only one team at a time is trying to score, and it has been studied on a large scale for decades. Football is a low scoring game in which territory is regularly gained and surrendered, and it is difficult to record who does what. Football statisticians used to focus on goals scored and conceded in order to make predictions.
Predicting the outcomes of matches is still done using variations of this method. Italy will beat England in the Euro 2020 international tournament, thanks to a mathematical model developed by epidemiologists at the University of Oxford. Six of the eight quarter-finalists were correctly identified.
It is not uncommon for such success to occur. Matthew Penn is a PhD student at Oxford who developed the Euro 2020 model.
He says, "You want to give each team an offensive and a defensive strength, and you work that out from the total number of goals that each team has scored." It's easy to predict each match once you have a large set of equations to solve for the two sets of strengths. According to Penn's model, Belgium has the best chance of winning the World Cup, followed by Brazil.
Information about events on the pitch and how players influenced them is of interest to coaches.
This type of information has been recorded by football analysts. Charles Reep, a former Royal Air Force accountant, spent a lot of his time in the 1950s watching matches in England and making basic observations. Reep used his data to propose strategies and tactics. He helped to create a playing style that frowned on sideways passes and won three league titles in five years at Wolves.
More than a decade ago, most top clubs and national teams began to use data analysts because of modern technology. Its use goes all the way into the football pyramid. Penn works as a part-time data analyst for Oxford City, a semi- professional club that plays in the National League South, the sixth tier of the English system.
Many analysts attribute some of the recent success of the London club to an in-house algorithm that rates players across different leagues. A model that can assess whether a player's actions on the pitch make a goal more likely has been created by the data team at the club. In a partnership with FC Barcelona, sports scientists at the University of Lisbon, Portugal, published an analysis of how long opportunities last for different types of pass.
Pre- match reports are the most useful thing that we do at Oxford City. We look at the attributes of the players for the other team, and then produce some graphs to show how they are playing. I will suggest some changes to the game. Penn noticed that the left back had poor heading statistics. He says the idea was to have the big man on the right side of the pitch. The game ended in Oxford's favor.
The naked eye of an experienced scout can see that. Penn believes that the data will be less biased than someone's opinion.
The clubs don't have to make the data themselves. They can buy the information from commercial companies that film matches to record the outcome of major in-game events. Computer vision is a type of artificial intelligence that is used to record such data. Each player's pass-completion rate is often included with the data.
The kind of detailed analysis not usually available in the raw data supplied was used to develop a model to assess the passing strength of all players.
It's important to take into account what kind of pass they try. Marks says that you can't just take the completion ratio. If you know an opponent is passing well from a certain area of the pitch, then you know what to look out for.
After working at Microsoft as a data analyst, Ramineni moved to Seattle to work for the MLS team. One of his first tasks was to use the gps data on the players to improve their training. It's possible that the training was too much or too little. To prevent injuries, you do that.
Is it possible that it worked? When we applied the methods, we had good seasons. I have no idea. If an injury doesn't happen, it's hard to quantify.
There is no control experiment to check on efficacy when it comes to the role of data in sports success. The coaches at Seattle were willing to listen to Ramineni's analyses when judging the strengths of players.
He says he was able to talk to players directly. It isn't the same at other clubs The coach doesn't interface with the data guy
When players don't have the ball, analysts pay attention.
According to Ramineni, one of the most important things to know about football is the player's off-ball skills.
It requires dedicated cameras that don't just track the main action, but also keep an eye on players who aren't directly involved and tag their locations 25 times a second. Ramineni says that companies that supplied this kind of data signed exclusive deals with national league.
He wouldn't know their off-the-ball metrics if he were scouting an international player from South America or Europe.
Data scientists are using technology to predict injuries.
In recent years, a more powerful technique has emerged that uses artificial intelligence to predict the movements of players in filmed matches, even when they are not captured directly by cameras Data companies can use broadcast footage of games to produce comprehensive on- and off-the-ball analysis for players anywhere in the world.
There is a partnership between the researchers at DeepMind and the data team at the football club.
Ian Graham quit his job as a PhD student at the University of Cambridge to work in football statistics and asked about tactics and counterfactuals.
The model can create thousands of simulations about what could have happened instead. You can talk about how well an attacking move went during that time.
The publishing of this work was an uncomfortable step for the club because they don't give details on what they are doing or how well it works. It was one of the conditions of working with Deep Mind.
We don't have the resources that we need to build our own models, and that's becauseLiverpool has one of the biggest and most-developed departments in football. He's reassured that no other club can.
Graham doesn't want to take credit for success on the field. He says that football teams lose when they shouldn't lose and win when they shouldn't win. Our analysis shows that we played well when the team is doing poorly. If we play like that, we will win a lot of games.
The off-camera modelling work is the first step towards creating a virtual assistant coach that uses real-time data to guide decision-making in football and other sports. He says that it's possible for the artificial intelligence to suggest a change in formation that would do better.
Tuyls says that the approach could be useful away from the pitch for tasks such as modelling the trajectory of self-driving cars and pedestrians.
What are we going to do next? The experts in football data stress the need for more research. Sarah Rudd was a data scientist at Microsoft for nearly a decade before she left the company last year to work for the London club.
She said it would be great to have that level of data. We haven't figured out how to derive insights from things that aren't being measured.
Data that shows players orientation and how they shift their weight is the next advancement. Rudd says that the tracking data isn't at the right level. You are not yet picking up that small stutter step or shift in weight that a player does to throw the defender off balance.
It is possible to be bamboozled by incomplete knowledge of a player's position The model might say that the player did a bad thing because he should have started running. He was lying on the floor when he tripped over.
How have numbers affected football?
Ramineni thinks recruitment is the best place to get a bang for your buck. When a team gets a free kick after play is paused, that's another area.
Data analysis shows that players shouldn't shoot when they're far away from the goal. The distance from where players have taken shots was higher a decade ago. Data-analytics people have begun asking why you are shooting from there. There is only a 2% chance.
Statistics showed that most of the crosses were pointless, so many teams now discourage players from trying them.
As the amount of data grows, so will the job opportunities. There is no going back on the data footprints.
Nature 611, 444-446.
The article is titled: "D41586- 022-03698-1."