“Everything is a number”
Valeryi Lobanovskyi, football coach
In 1973 Valeriy Lobanovskyi became coach for Dynamo Kyiv, the main football team of Ukrainian capital. He immediately changed the history of football with two rather unusual requests: a computer and a statistician. The bizarre request of a computer, a mysterious and rare machine at that time, attracted the attention of the Soviet spy agency, the KGB, which put it under control. “Everything is a number”, Lobanovskyi prophesied, training his team with the help of statistician Anatoly Zelentsov and the computing power of his supervised computer. He was searching for a scientific method to define tactics and schemes, teach players how to cover the pitch and pass the ball, and finally beat the opponents with the geometry of numbers. He approached football from a mechanistic point of view, using automatic computations and statistical analysis to understand the patterns of success in football games. “A team that commits errors in no more than 15 to 18 per cent of his actions is unbeatable”, is one of the scientific principle at the basis of Lobanovskyi’s way of coaching. One of the first in football history, he conceived a sort of total football, where universal players are able to cover any role on the pitch. The revolutionary ideas of “the colonel”, as Lobanovskyi was named for the grueling workouts and the discipline he imposed to his players, led the Dynamo Kyiv machine to the most senior European football: two UEFA cup winners’ cups, one UEFA super cup, two Golden balls for the team’s most representative players (Blokhin and Belanov). As coach for the Soviet Union national team, the colonel reached the final of European Cup, where the Soviet geometry of his numbers was defeated by the Dutch magic of the unpredictable Tulips.
Two decades before, when computers were still prototypes and Big Data just science fiction, British accountant Charles Reep was used to fill his notebooks with data about football games, creating the field of sports analytics. From 1950 to 1996 he collected by hand more than 2,000 football games, annotating all the main events of the pitch such as all the passes, their length, direction, height, outcome and even the position on the pitch at which the pass originated and ended. He then analyzed the data collected, rigorously by hand, with the help of professional statisticians. Thanks to Charles Reep for the first time several aspects of football were found to follow strong and stable numerical patterns. He discovered for example that, on average, teams score with roughly one of every nine shots they take, that the 30 per cent of all regained possessions in the opponent’s penalty area led to shots on goal, that about half of all goals came from those same regained possessions. His most remarkable observation is that the probability of loosing the ball increases with the number of consecutive passes. This led Reep to define the so-called “long ball theory”, a demonstration of the efficacy of counterattack: the most efficient way to score is to bring the ball close to the opponent’s goal as soon as possible. His seminal findings, published in two papers on the Journal of Royal Statistical Society, inspired the English football school of the sixties and seventies.
Nowadays, football data are more abundant, rich and precise then they were at the time of Charles Reep. The datafication of life has started to infect sports, and football analytics has evolved in an amazing way thanks to the modern sensing technologies: during every game, teams of modern Reeps support semi-automated sensing technologies in providing high-fidelity data streams. Even though data and computers cannot do the manager’s job, they help coaches and managers to understand the complex behavior of players on the pitch and to build a more successful team. Take the case of Midtjylland, a Danish football club founded in 1999. Manager Rasmus Ankersen and owner Matthew Benham used data and mathematics to develop a top-secret “algebraic scouting algorithm” to select the right players during the football market and locate them in the right position on the pitch. This algorithm made the young club to win (against any forecast) their first Danish championship and then to access the prestigious European Champions League. The market choices of Midtjylland were guided by algorithmic data analysis and a profound knowledge of the patterns of football. When they bought Tim Sparv from a team in the German second division they knew that, despite he was a defender, he rarely played on tackles. In other words, Sparv complied with the “Maldini principle”: a good defender (like Sparv was found to be) does not play on tackles because he has a great sense of position, is able to rapidly evaluate the next step of the opponent, and hence does perform the right movement avoiding tackles and faults. Ankersen and Benham knew that, when evaluating the quality of a defender, counting statistics on tackles can be misleading. A lesson that sir Alex Ferguson, to his detriment, learned some years ago. In 2001 Manchester United’s coach looked at football statistics of his players and found that defender Stam was showing a tremendous drop in the number of tackles. Ferguson, ignoring the Maldini principle, wrongly thought that Stam was in decline and sold it to SS Lazio. It was a terrible mistake. Stam was not in decline at all and played high level football for another six years in the Italian league. Stam’s tendency to avoid tackles, indeed, was not a sign of decadence but a sign of maturity.
Today’s teams, at least the strongest and richest ones, are starting to build a data science department to realize and improve the lessons of Reep and Lobanovskyi. Every game of the team is analyzed by sports data scientists in every possible dimension, using the Big Data dearly bought from companies like Opta and Prozone. In the past, coaching was dominated by insight and experience. The great coaches in football history — from Helenio Herrera to Arrigo Sacchi and Pep Guardiola — were capable of capturing profound insights about the patterns of football and to implement them in new revolutionary successful strategies. Now the situation is changing. Besides a team of players, a team of sports analysts is emerging in football clubs, with the purpose of mining from data invisible schemes, tactics and winning patterns. Teams that use data science to improve team’s and players’ performance are gaining an enormous competitive advantage against their peers. Of course, the new emerging data-based football coaching needs coaches and managers able to understand and interpret data analysis, graphics, results. Tomorrow’s coaches will be rather similar to a mix of Reep and Lobanovskyi, and football fans already know that experience and insight are useful but not sufficient features anymore. Their favorite team’s next coach should be also a good data scientist.
Article by Luca Pappalardo and Paolo Cintia