Sport is a field of interest where humans have always tried to push the limit. Every day a new record is broken and the frontier of performance is redefined. Technology helps in this direction. Technology as well is evolving. Artificial Intelligence in sport is here to help both coaches and athletes to enhance physic and psychologic performance. This is happening both at individual and team level.
Availability of data made this possible. Human memory is programmed to remember the substance and meaning of events, not the details. Machines can record and analyze infinite details at speed light. Cameras, videos and sensors applied to body and equipment can collect a quantity of information much higher than a coach and any assistant can do. And can make sense of all those data, through pattern recognition. In real time.
The main reasons are relatively simple. Win and last longer. Analysis can help to improve performance and make strategy and playing tactics more effective. This leads to better results. But a scientific approach to training, playing and managing injuries can also help athletes to increase their longevity on the field. Both items have a clear link with profitability and I bet nobody is really surprised about it. This is one of the main drivers for artificial intelligence to conquer new territories in sport and become an established practice for players and teams. Artificial intelligence in sport is here to stay.
At this stage I think we need to do a bit more clarity on what we are really speaking about and why it’s different from statistics. Simple event statistics do not capture the complex aspects of a game. Machine learning approaches can do more than traditional statistical techniques, because they can make sense of random patterns in large data sets. Algorithms can be used to identify potentially complex yet meaningful patterns in the data. When a pattern is recognized, it can help to predict future events or simply to adapt current behavior (like a tactic or a decision on the field). Somebody say this is just brute computational power. I disagree. Computational power is about extracting the information in an automated way, by looking at a match or the video of a match; while machine learning is a form of artificial intelligence, where computers are able to learn without being explicitly programmed by a human operator. On the other side, I agree that’s not general artificial intelligence, because none of those algorithms are useful outside their narrow field of knowledge.
This does not mean we can underrate the role of artificial intelligence in sports, because it’s becoming huge and will have implications on how we enjoy and consume the sport “product”.
Artificial intelligence in sport: analyze in real time
Analyzing in real time is the first part of the story. Machines have the ability to ingest a variety of data sources and large quantities of data. We call them “training data”, because they help the machine to learn. And usually, the more they are, the better. Cameras and sensors collect the data. Then scientists describe what has happened during the activity by applying some complex reasoning algorithms which allow them to determine the tactics and types of play that are happening on the scene. Actions will be different depending upon what sport we are watching and peculiar to each sport. Then after a number of hours of “watching”, the machine will distill the patterns and their ability to impact the final result. But what’s exactly the benefit? Objectivity. Coach decisions are affected by what he sees (and doesn’t see), his background, knowledge, the context or even the feelings of that specific moment.
Elite team are all moving in this direction and investigating how data analysis can help improve their game. In training sessions, players wear GPS trackers, acceleration sensors and heart rate monitors to analyze their training performance and optimize their preparation. Professional labs also match those data with other data about their players, like sleeping habits, diet and many other health kpi’s to get an unparalleled view of their status.
When it comes to the match, artificial intelligence in sport is even more comprehensive. Bayern Munich recently partnered with German software giant SAP to gain detailed performance assessments after each match. Today, all Premier League football stadiums in the UK are equipped with a set of digital cameras that track every player on the pitch. Companies like Prozone analyze all data, provide comprehensive insights across all phases of the performance analysis cycle, offer a post-match analysis platform and customize the feedback to each client. Another giant is ChyronHego. Their system is installed in over 125 stadiums and is used in more than 2,000 matches per year by the Premier League, Bundesliga and Spanish La Liga. At 25 times per second, the system generates live, accurate X, Y and Z coordinates for every viewable object, including players, referees and even the ball. The data provides insight for coaches to evaluate player performance and track metrics such as distance run, speeds, stamina, pass completion, team formations, etc. An Australian company named Catapult Sports has developed a player’s tracking system based upon Global Navigation Satellite System (yes, you understand correctly) and they work with more than 450 teams worldwide, including Chelsea, Real Madrid and Brazilian national team.
The FIFA reaction, up to today, is that players can adopt wearables or tracking systems during the match, only if information is not available to coaches during the match. Which is clearly fair and avoids that, the one having the best algorithm has an advantage, but it’s not going to last. Sports based on high technology equipment can already use data without any issue (for example in racing). I expect that, with some excuses about the player’s health on the field, also this barrier will fall down. And, anyway, when all teams will have a technology available, it will make no sense to preclude them from using it.
|If you like my post, please share it, thank you|
Artificial Intelligence in sport: Making the computer the coach
Making the computer the coach means quantifying the sport and predicting what is the best decision on the pitch, both in terms of team schemes and individual players’ action.
A great article of the Newscientist explains it well: “A detailed analysis of the passing strategies of 20 teams in the Spanish football league during the 2013-14 season has given a unique insight into how they play. Information on the players involved, their pitch co-ordinates, the distance between them and the time taken for each pass was studied by a computer, which had analyzed video footage of games. It pored over 300,000 individual passes made across the entire season and identified hundreds of patterns used by the teams. It also looked at whether they occurred in more than one game. Sure enough, the algorithm revealed that Barcelona and Real Madrid had more than 100 recurring passing patterns, 151 and 180 respectively, and retained possession in their own half. But there were surprises, too. Atletico Madrid, which won the league that season, had just 31 recurring patterns. Stefan Szymanski, a professor of sport management at the University of Michigan, says this lack of predictable play could be why Atletico Madrid had such a successful season.”
And that’s the point. Machine learning can tell a coach, young or veteran, what’s the best decision in every situation. But there is a difference between what has statistically worked well and what is necessary to be successful in a specific moment. And it’s the human factor. Creativity and unpredictability. Humans can do the unexpected to crack the AI suggestion in the hands of the opponent. So I agree that having an AI companion can help to make sense of an opponent’s playing style, but even if a player has a computer telling him what might statistically be the best play for success, there’s still a possibility that could change in the moment. The authenticity, the instincts, the risk-taking are all crucial elements that an artificial intelligence in sport cannot entirely replace. And this is even truer for team games, where the combinations of player actions and reactions are extremely numerous.
Without being naïve it’s clear that teams are trying to find solutions to enhance their performance and to extract that extra 1% from data. A small increase in performance can determine the difference between winners and losers, but when every team will have a good algorithm for predictions and players will understand that the unexpected is a good decision, the benefits of an artificial intelligence in sport will be diluted.
Newsletter: because there’s more than artificial intelligence in sport here!
The Futurist Hub Newsletter is the greatest thing after the Big Bang. Once per month, only the news, free of spam. And with a free ebook as a bonus.