[ad_1]
March 1950, An RAF wing commander and a trained accountant named Charles Reep stared at the numbers in football. Reep, who sparked interest in the sport in the 1930s and was fascinated by Herbert Chapman’s pioneering Arsenal team, returned from World War II to find out that the tactical revolution he had seen before had come to a halt.
Eventually, during the break during a black Three Division match between Swindon Town and Bristol City, while witnessing numerous attacks, Reepen’s patience was exhausted. He grabbed a notebook and a pencil and began to get angry at what had happened on the field: he began counting passes and shots in one of the first systematic attempts to use data to analyze football.
Seven decades later, the data revolution has reached the villages — fans know it well xG and net spending, and the best teams receive statistics to directly benefit college doctoral students. Now, Liverpool’s first league champions Liverpool have teamed up with DeepMind to explore the use of artificial intelligence in the football world. The work of researchers from both organizations, published today Journal of Artificial Intelligence Research, explains some potential applications.
“It’s the right time,” says DeepMind AI researcher and one of the paper’s lead authors, Karl Tuyls. DeepMind’s collaboration in Liverpool stemmed from his role at the city’s university. (Founder of DeepMind Demis Hassabis He is also a Liverpool fan and was a research consultant.) The two teams met to discuss where AI can help footballers and coaches. Liverpool have also provided DeepMind with data on all the Premier League games the club has played from 2017 to 2019.
In recent years, the amount of data available in football has increased with the use of sensors, GPS trackers and computer visual algorithms to track the movement of both players and the ball. For football teams, AI provides a way for coaches to identify patterns they can’t; For DeepMind researchers, football offers a limited but challenging environment to test their algorithms. “A game like that [soccer] it’s very interesting because there are a lot of agents, there are aspects of competition and cooperation, ”says Tuyls. Unlike chess or Go, football has its own uncertainty within it because it is played in the real world.
That doesn’t mean you can’t make predictions, but it’s an area where AI can be especially useful. The paper shows how a model can be trained in data about a particular team and squad to predict how players will react in a given situation: if a long ball hits the right channel against Manchester City, for example, Kyle Walker will move in a certain direction, John Stones something else while he can do it.
This is known as “ghosting” —alternative routing overlaps with what actually happened, as in a video game — and has different applications. For example, the effects of a tactical change or how an opponent can play can be used to get a player out of a key injury. These are things that coaches would notice for themselves, and Tuyls stressed that the goal is not to design tools to replace them. “There’s a lot of data, a lot to digest, and it’s not so easy to handle that mass of data,” he says. “We’re trying to build support technology.”
As part of the document, the researchers also analyzed more than 12,000 penalty kicks across Europe in recent seasons — classifying players into groups based on their style of play, and then using that information to make predictions about where they were most. they are likely to throw a penalty and see if they can score. Strikers, for example, were more likely than midfielders to play in the lower left corner – they took a more balanced approach and the data proved that the best strategy for penalty takers was perhaps not surprising, as they were throwing at their strongest side.
[ad_2]
Source link
