How is AI and data analytics improving performance at the world’s elite clubs?
Football at the elite level is known for its innovation, intensity and tempo, and yet of all of the world’s top sports, its adoption of technology has been perhaps the slowest. VAR and goal line technology have only recently appeared in the game, and while sports like Basketball, Baseball, Cricket and NFL have embraced the possibilities that data analysis creates, football clubs are only beginning to scratch the surface of its applications in the beautiful game.
It is in the context of scouting and transfers that data analysis is best known in the footballing world. AI technologies help clubs analyse the candidates they have in mind in a faster manner, but they can also highlight to staff previously unheralded players who fit certain criteria or possess key attributes. London-based Brentford are the most popular example of a club that use a data-driven approach in scouting to punch above their weight. Their deployment of data analysis has been a great success, helping them sustain a model based on recruiting players from lower value leagues who outperform several metrics, to then sell them on for considerable profit.
While the use of technology in scouting is hyper-relevant in a footballing world that has been financially impacted by Covid-19, it is in tactical analysis that the most exciting strides are currently being made. Several clubs are global leaders in the use of AI and data analytics as tools for tactical analysis, including Barcelona, Liverpool, Seattle Sounders, Manchester City, Arsenal and Toronto. However, the tactical revolution is still taking place at the lower levels, and for many clubs the effective use of AI is not common practice yet. For those who do get it right, the use of AI and data can give clubs a better understanding of their opponents and maximise the value of their own playing squads by improving performance. At a time when massive transfers fees could be on the wane, developing players will become increasingly important to clubs, and the technology now exists to help them do this.
Football clubs are time-poor, and saving time is still the main advantage of using these AI technologies at the elite level. However, there are increasingly more ways the technology can assist teams to get an edge over their opponents. AI and data analytics can provide new intelligence or improve existing operational efficiency. For example, using AI models to identify the key ways in which an opponent attacks has two clear advantages: On the one hand, it can save you from watching hours of videos, but on the other, it can also help you identify something that a human eye might have missed when visually inspecting the games. Tracking and event data are sophisticated enough that clubs can now track every moment their defenders move too create a distance from each other, or which players are pressing the opposition ineffectively.
Whether it is improving operational efficiency or crafting new intelligence, incorporating AI and data science into the workflow of clubs is about winning more games. However, to win you don’t only need to be great at tactical analysis, but you also need to be great at communicating your findings to the players. New generations of players are increasingly receptive to visual learning, and more engaging data-driven presentations are becoming the norm at clubs to help players digest key information in the build up to games.
Stockport County, one of the clubs we work with, are a prime example. They have combined their tactical analysis with our technologies to devastating effect in a recent FA Cup performance against League one outfit Rochdale. Their performance analyst, Sean O’Callaghan, revealed the opposition’s defence liked to press high and close the opposition’s space, meaning that Rochdale’s full backs were often caught high up the pitch. Having identified a weakness to exploit, Stockport’s winning goal – against a team several leagues higher up in the pyramid – came from a ball that was played over the full back.
While we are starting to see the usage of AI driven tactical analysis in soccer, what we are looking at is just the tip of the iceberg. To have an intuition at what the next years are going to look like, we can check what already happened in other domains.
In chess, for example, players use chess engines to prepare for opponents with specific playing styles and has the potential to be the next big breakthrough in football. While the progress in football analytics so far has been done mostly by computing models based on historical data, there has been very little progress or research done on simulating the game of football. Drawing inspiration from chess, that could change dramatically in the future. Clubs could be simulating different systems and even different players against one another in the build up to a fixture.
In July 2019, Google Research released the Google Research Football Environment (GRFE), a novel RL environment where agents aim to master the world’s most popular sport. Manchester City and Google Research then announced a Kaggle competition using the GRFE to explore the ability of artificially intelligent agents to play in complex soccer environments, with the goal of opening up the environment for the community to foster progress in the generation and understanding of the game. The fact that the game of football can be considered a Markovian system makes it a great to apply reinforcement learning methods on, which is an extremely exciting future!
While the pace at which elite clubs are embracing many of these technologies is developing quickly, this type of analysis is still outside of the grasps of the vast majority of clubs. One of the main reasons this is happening is because of the cost of tracking data. If we want to reach the next frontier on tactical analysis and innovation in the field, we need to find a way to provide clubs at all levels of the game with more affordable data. We have a solution for that: our new Automated Tracking Data which is now at an early access stage to some of our customers and will be available to all in the near future.
Some of the most useful analysis clubs do, like tracking space between defensive lines, analysing a team’s press or their overall shape and philosophy, computing passes options or pitch control models, don’t rely on the players identity. You need to know where players stand, but not who is who. ATD gives you exactly that: the positions of each player and the ball in each frame. This can then give analysts at all levels things like tactical shape identification, similarity search, pitch control, EPV models, opposition analysis and scouting.
For analytics to take the next step in its football journey, it’s obvious the software must become more democratised and accessible to everyone. We know that there is a vast community of people out there who can do amazing work if they have access to the right technology. This is why we’ve decided to release PLAY by Metrica Sports’ Free Plan, which is free to download to anybody, anywhere in the world.
Now the next generation of analysts can have unprecedented access to the technology being used at the elite level of the game, and the opportunity to use that same technology themselves, whether that is to create analysis of their own matches or footage from professional games. Why shouldn’t some friends playing in the park be able to use the same tools as the international superstars they idolise every week? Now they can.
Bruno Dagnino is the CTO of Metrica Sports, a leading data and video analysis platform that has provided cutting edge solutions to elite football clubs since 2014. Aspiring analysts, players and coaches, can download Metrica for free and try their hand at video and data analysis.
