“Artificial Intelligence will not replace analysts because it should never make decisions”
Joan Fabregat, Project Data Manager at FC Barcelona, has chaired a discussion between Ivan Valle, Video Analysis and Sport Performance Systems Manager at FC Barcelona, Joe Pavitt, from IBM, David Garcia, atsystems data engineer, and Kristine Stewart, from the World Economic Forum, on decision-making based on data and Artificial Intelligence.
Valle described the progress that has been made at Barça in this area. The club generates 120 terabytes of video a year and in association with the Pixellot company records football on the youth and first team training grounds. “We record absolutely everything,” he claims. The purpose is to assist everyone from coaches and scouts to scientists with their decision-making.
The challenges faced by the club, he revealed, include obtaining information in real time and centralising all data. Currently, Barça serves as an experimental laboratory for sports data and is providing young athletes with very useful tools to boost their performance.
In recent years, Pavitt has been specialising in applications of Artificial Intelligence to sports performance analysis and broadcasting. His main objective is to define an interface to present data. Easy access to information is the key. A natural language must be used that enables specialists from different areas to consult data with ease.
These models have already been developed in rugby with an app that provides the biggest plays from a game immediately after it has finished, which are mainly used by the players themselves. Many of these are not professionals, but have regular day jobs and don’t have the time to analyse their game. Therein lies the main utility of these technological developments: freeing up analysts’ time so they can spend it analysing, rather than extracting and preparing data.
Garcia stressed how the use of Artificial Intelligence would help an athlete to boost their performance. They can use this app to assess the likelihood, type and duration of an injury, or even the chances of success in a competition. The only problem is with the quality of the data and where it might end up being published.
For Stewart, the pandemic has sparked a revolution in sports broadcasting, which has naturally involved Artificial Intelligence. The work that has been done to add crowd noises and make it seem like there are fans inside the stadium when games have been played behind closed doors has not been an isolated circumstance, but rather has shown the way forward.
“Even if you diversify your revenue, you should never forget what your main business is”
FC Barcelona Ambassador to the United States, Steve Gera, has spoken with Shana Ferguson, Head of USA Swimming’s Commercial Department, Larry Freedman, President of Los Angeles FC, and Eric Winston, of OneTeam Partners, about alternative avenues for monetisation and new business in the world of sports.
Ferguson explained that as a government-linked organisation that prepares athletes for the Olympic Games and World Championships, its alternative revenue streams are limited. Its budget mainly comes from the fees paid by its 400,000 members, who are between 6 and 18 years of age, and the donations they receive. However, she says that “every dollar that comes in, goes out for the members.” That is to say, everything they collect is invested in the development of swimming.
Right now the pandemic is hurting them badly. Pools have been closed and children are not paying fees to join the organization. There have been no training sessions, and there have been no competitions. When the virus hit hardest, absolutely everything was stopped, but even now that 80% of the clubs have reopened, fewer children are signing up because their parents are scared. Hence advertising revenue has dropped too. To find a way around the issue, they have tried holding virtual competitions, but as they have a long-term sponsorship strategy, they decided not to accept any new offers from advertisers. “Maybe we were too good,” she laments. “Because we have still had to go on paying the costs of all our employees and tournaments.”
Los Angeles FC’s greatest asset is the stadium. Apart from football matches, for which they receive revenue from ticket sales, hospitality, gastropub, merchandising and parking, they also make money from conferences, concerts, meetings, TV shows, eSports and festivals of all kinds. With the pandemic, matches were broadcast on TV and the empty seats were covered with advertising. They also opened other lines of business to partner with other companies, especially when seeing how Adidas sold black and gold sneakers in a matter of minutes, but the point he really wanted to drive home is that: “You can never forget what your main business is.”
Winston, for his part, explained how his company helps athletes to monetise their image. In his point of view, athletes are not being attributed the market value they deserve. Gera asked him to explain why his firm is famed for doing business outside of traditional margins and Winston replied that athletes can divert the use of their image to businesses outside of sports. He has cited a deal he is working on with a firm that is developing an Uber for hair extensions, which are very expensive, and what a great investment opportunity that is for African-American athletes. Because he believes that we must move on from the sponsorship era and focus more on joint investments, whereby the athlete and the firm become partners.
“Why is tennis lagging behind in data analysis?”
In 2004, a US Open game between Serena Williams and Jennifer Capriati involved several controversial plays. Although there was video, there were no Hawk-Eye systems to help the line judges decide on questionable balls. Williams’ anger that day made history, and since then electronic review has become a vital part of the game. Just two years later, the US Open was the first Grand Slam to use it. However, Stephanie Kovalchik, a data scientist at Zeus Analytics, believes that there is still a long way to go and no shortage of obstacles.
Although tennis was one of the first sports to use tracking to analyse ball trajectory and player position, such information does not add much to what a viewer can already see on TV. For Kovalchik, second-screen developments in tennis broadcasts give more or less the information that can already be seen on television. There have been attempts to offer more advanced analysis, but ultimately “it has done no more than calculate where the ball will land.”
The fact is, says Kovalchik, there is no widespread awareness that data could revolutionise tennis as much as is already happening in other sports like football. We need to think outside the box and introduce analysis that provides more than mere descriptive statistics and that truly tells us why someone has won or lost.
The paradox is that tennis has such a strong tradition of data analysis. She cites Bruce Old as one of the pioneers, author in the 1960s of several books on tactics. Unfortunately, she doesn’t think many modern-day coaches have even heard of him. The same occurred with the academic studies on tennis of the 1970s, which were already predicting phenomena that would not appear in the game until around 2010, one such example being the research into serving strategy in tennis by SL George in 1973.
However, there is currently a divide in the use of data. There is no good access. There isn’t even an official supplier. The problem is that data is collected independently by each tournament, just like they do with players. Additionally, tournaments are interested in attracting the best players and they earn revenue from broadcasts, ticket sales, and advertising. Improving the game is not one of their priorities and the data on more than 80 tournaments is collected individually, and afterwards they do not share it with each other. That’s why tennis data has to be sought outside the official channels, and generally from volunteers.
There is room for the players to improve and for the game to develop, but none of this is being exploited, laments Kovalchik. For example, simply by counting drop shots, it can be observed that their use been increasing statistically in the game. If tennis data was free-access and centralised, more phenomena like this could be identified.
“For twelve years, Barça has had most possession in 887 of 902 games”
FC Barcelona’s philosophy is based around the ball, and passing is the way the players communicate with each other. Something as simple as this, the so-called natural game, has infinite possibilities and is something, says Isaac Guerrero, that a whole lifetime can be spent studying. However, his talk today focused on merely describing the main elements behind the way the club has yielded such stunning stats in terms of dominance of possession: in 887 of 902 games it has played in the last twelve years, Barça has had the most possession.
Guerrero differentiates between ‘model’ and ‘idea’. A model would be a pattern that players are obliged to copy. An idea, however, is something to be adapted to the context of each game, giving the player freedom to operate within the framework of a group plan. In Barça’s case, it is the ball that is responsible for organising the game.
The initial premise is that it is the ball that has to find the players, not the other way around. He quoted Seydou Keita, the first signing of the Guardiola era, who described the way his new team played as “amazing”. He said that at all the clubs where he had been before, he had to run all over the place to get the ball, but at Barça he just had to stand his ground and the ball would come to him. Guerrero added that the short game also benefits ball recovery because the player only has to move two or three metres to do so.
Both concepts are part of locational football, an evolution of what was known two decades ago as positional play. The common criticism of that approach, Guerrero recalls, is that the team “does not advance”, but he feels that was because teams had yet to get to the ‘heart’ of the idea. Barça players continually reorganise themselves, they are constantly changing location. By moving the ball from one place to another, passing lines eventually emerge until the goal finally comes into sight, the final link in the chain. The killer pass.
Since Rinus Michels, the club has been continuously developing this philosophy. The most important steps were taken under Johan Cruyff, but before him Laureano Ruiz had already introduced the concept to the Barça youth teams. It was also present in the Menotti era of the eighties. Later on, in the nineties, Louis Van Gaal made a major contribution by introducing his own interpretations of the system. After that, Rijkaard, Guardiola, Tito, Luis Enrique and others have made history with the same basic system, and every coach since, from Valverde to Setién, has followed suit.
However, playing to pass is not the same as passing to play. The important thing is for the whole team to be connected. It is not a matter of some attacking and others defending; there has to be a connection between them all. Similarly, actions are not about balancing risk-benefit ratios. This model is not about goals, it’s about the ball, and is a method that takes the pressure off the players. The idea is that when close to goal you do the same things that you do in parts of the pitch that are viewed less as hotspots. But the sine qua non at the root of this system is fostering player autonomy.
Questions are sometimes more important than answers
Dr. Asker Jeukendrup, researcher and professor at Loughborough University and considered one of the best sports nutritionists in the world, has spoken about the importance of asking good questions and their relevance when working with athletes. As he says: “if we are not asking the right questions, then the answers might not be relevant.” Nutritionists need to solve athletes’ real problems, so the questions they ask must be clinically relevant. But we commonly encounter research that exaggerates the effects in order to study proofs of concept.
For example, he discussed the evidence showing that combining caffeine (in doses of 10 milligrams per kilogram) with carbohydrates increases the intestinal absorption of glucose. There are indeed studies that have shown this, but it would not make sense to recommend such a dose to athletes due to its possible side effects. So, although an improvement in intestinal absorption of glucose has been demonstrated when combined with caffeine, it is not easy to transfer this knowledge into clinical practice.
Similarly, other studies have described caffeine’s capacity to increase fat oxidation. “There is indeed research that has found that caffeine can increase fat oxidation by 27%, which would translate into an absolute difference of 5 grams per hour. This means that, to achieve a loss of 1 kg of fat, we would have to do about 200 hours on a bike. And that would only be true if we assume that fat oxidation results in weight loss, which it does not. We need a negative energy balance.” So, although the media often heralds grandiose conclusions to certain research, we need to interpret the results and assess the possible applications to clinical practice.
In the final part of the presentation, the researcher detailed strategies that can improve gastrointestinal symptoms, explaining the differences that take place in the stomach and intestine depending on whether low-releasing or fast-releasing carbohydrates are taken. “The more carbohydrates there are in the intestine, the greater the risk of gastrointestinal problems” he said. “Because small intakes of fast-releasing carbohydrates are the ones that put the least pressure on the digestive tract, they are likely to be the best option for the stomach and intestines.”
Dr. Jeukendrup ended by speaking about the importance of using the Socratic Method when guiding the athlete with decision-making. It is the athlete who should find the ultimate solution to the problem raised by the nutritionist through a process of close communication. “We need to ask more questions when working with athletes rather than immediately giving them our knowledge or advice” he says. “We should try to understand the athlete a little better. A trusting relationship needs to be built, and that will also let us give more effective advice.”
“Coaches and data specialists need to coordinate criteria so they are both speaking about the same thing”
Alex Thomas, Data Architect for the English Football Association (FA) has explained how he organises the information that is collected from English football. Data science has changed a lot in the last eight years, he says. Concepts like expected goals are no longer such a mystery, which is pushing scientists toward more abstract ideas.
The organised implementation of data philosophy requires one essential principle: clear information. There is nothing more important about a strategy than the definition of each concept in order to collect each data in an agreed and centralised manner.
His premise is to define beforehand what type of data serves whom. In football, this information is used by anyone from psychologists to nutritionists, as well as coaches and scouts. They use it to obtain all the patterns needed to understand the game and the reference points to order the information. Isolated information on a video clip, for example, is useless. You need to know if it is a final, a friendly, if the team is winning or losing, the minute, score, etc… Only then will the file make sense.
His procedure is therefore based on data labelling, which essentially means watching the images and adding categories to them. If we talk about a pass, we need to know with which leg, the distance, if it reaches its target, if it is in the evening or afternoon, if it is hot or cold, if it is competitive or a friendly, and so on.
So agreement on the terminology is very important. When a coach talks about pressure, Thomas wants computer scientists to understand the same meaning of the term. Their criteria need to be coordinated to ensure they are always speaking the same language. Those definitions are the biggest challenge faced by the analysts who code these games.
Thomas says it is also important to establish these criteria for the sake of in-house data compilation. Companies receive large amounts of data but very little of that information is useful. If they collect their own data, he believes, there will be less confusion and it will be more accurate.
How to incorporate complexity into daily sports practice
Dr. Samuel Robertson, a professor at Victoria University in Melbourne, has spoken about ways to incorporate complexity into one’s day-to-day training routine. Although many sports like baseball still mainly work with basic ‘bulk’ workouts, which may lack specifics but nevertheless seem to produce improvements, Dr. Robertson claims that “the growing scientific literature shows that complex practice is more efficient and more effective than non-specific practice.”
As to how the processes and effects of training are to be measured, the biggest problem lies in knowing that although practice is complex, the measurement methods are not always so. Also, we frequently fail to measure everything that really matters in training or in competition, and instead focus on parameters that are not really relevant. The researcher therefore insists that we need to measure the right way in order to design programs, assess the players and their progression, and handle the different situations that can arise around the game. This is an important point because it is far too common for us “not to know why an athlete does something right or wrong in sport because we aren’t measuring what really matters.”
Technology is going to help us on the path towards this paradigm shift. Without it and without the right measurement methods, we will not be a relevant part of player improvement. Professor Robertson insists that we should not be doing things that we are naturally bad at. “We often try to measure as human beings, using our memories and recollections, and we are not especially good at doing that” he says. “That is why it is very important to measure objectively using data and models.”
Finally, he stressed that it is far too common for coaches to believe that their work counts above all else and is the most important thing, “but we must be aware that our role is subordinated to improving the athlete’s performance.” To do this, we must make athletes a part of the process and include them in the design of daily practice, something that will lead them to reflect, an essential aspect of getting to learn better.
How are hamstring injuries rehabilitated at FC Barcelona?
This presentation featured three FC Barcelona physiotherapists: Erola Madrigal, Juan Carlos Pérez and Xavier Linde explaining from an eminently practical point of view how players with hamstring injuries are rehabilitated, from the initial phases to the return to play.
Erola Madrigal began by presenting the case of a professional footballer who received surgical treatment after a hamstring injury. Erola described the different phases followed with this player week by week until he was able to start training with a fitness coach, from the first week when he was walking on two crutches through the progressive inclusion of exercises to regain his range of motion and strength. For example, we were shown how, in the third week, the player was able to start walking with one crutch, to work on electrostimulation of the quadriceps, and to walk in the pool. By the fifth week he was able to start walking and climbing stairs, and to perform exercises involving greater motor control (e.g. 90º squat and monster walk), by the seventh he was able to run on Alter-G (anti-gravity treadmills) and do single-leg supports with a load, and by the final weeks (9-11) eccentric and ballistic, plyometric and direction-change exercises could be performed before going on to work with the fitness coach. He also described the importance of controlling the rehabilitation process through electromyography and isometric dynamometry of the hip and knee muscles, and highlighted the importance of individualising the content and duration of each of these phases depending on each athlete.
After that, Juan Carlos Pérez explained how gym exercises can progress from analytical to functional work after a hamstring injury, the aim being for the athlete to later go on to work on the pitch. He presented tests that are used to evaluate the range of motion and functionality (including manual dynamometry and strength platforms to assess jumping), as well as various examples of exercises that can be included in a rehabilitation process, including isometric, concentric, eccentric and complementary exercises (for the quadriceps, adductors, gastrocnemius and soleus, glutes and core) and functional exercises (advanced strength work together with plyometrics, neuromuscular control and coordination, and learning to run again by doing skills exercises and Alter-G work).
Finally, Xavier Linde described how the return to play works at FC Barcelona following an injury. He explained the utility of sand exercises (which reduce loads at the tendon level while also stimulating the tendon in different directions and activating the muscles) and of running upwards and downwards during the early stages. He then described how the load should be phased and controlled during the return to play, starting with linear exercises and later going on to include acceleration, deceleration and changes of direction. He also stressed the importance of progressively increasing speed to regulate the stimulation of the hamstring muscles. For example, the number of meters travelled at high speed (>21 km/h) can be measured using GPS to control the load produced, and the ‘normal’ loads of that particular athlete must also be considered so that during the final stages of the rehabilitation process we will be as close as possible to those requirements.