From big data to artificial intelligence to predictive algorithms: how technological innovation assists busy athletes and coaches in Qatar
The 2022 World Cup has come to life – And if we refer to predictions it is impossible not to think of forecasting algorithms, data science: it is precisely these days, for example, the news that a team of scientists from the Alan Turing Institute has developed a forecasting model which, based on 100,000 simulations of World Cup matches, predicted that the Brazil national team will lift the World Cup on December 18 at the Lusail Stadium.
But science and technology are not only involved in these aspects, which perhaps interest bettors more than fans: today, in fact, science and technology are profoundly transforming football, supporting coaches and athletes ever more closely to develop better tactics and improve own performance.
Here’s how it will end
Let us begin with the possibly more “trivial” and certainly more antiquated side of World Cup predictions for 2022. As David Adams describes in a recent post published on the Nature website, statisticians who dealt with football for decades focused almost entirely on the amount of goals scored and conceded by a team and developing the best model to forecast them. Variants of these methods are still used to forecast match outcomes today: one, for example, assumes that the number of goals scored and surrendered is distributed around a given average value and was created by a team of epidemiologists from the University of Oxford.
It did us well, given that he predicted (correctly) that the Italian national team would beat the English national team in the final of the 2020 European Football Championships (more precisely, he predicted the greater probability of this with respect to goals scored and conceded); but not only that: he had also nailed six of the eight teams that reached the quarter-finals. “Basically, we want to arrive at attributing to each team an offensive and defensive ‘score’ – Matthew Penn, one of the developers of the model, explained to Nature – calculated starting from the total number of goals that each team scored and from the strength of their opponents: by inserting these parameters we obtain a set of equations to be solved to calculate the two scores, and it becomes relatively easy to make predictions for each game”.
It served us well, given that he correctly predicted that the Italian national team would defeat the English national team in the final of the 2020 European Football Championships (more precisely, he predicted the greater probability of this based on goals scored and conceded); not only that, but he also correctly predicted six of the eight teams that advanced to the quarter-finals. “Basically, we want to arrive at attributing to each team an offensive and defensive’score’ – Matthew Penn, one of the model’s developers, explained to Nature – calculated starting from the total number of goals scored by each team and the strength of their opponents: By inserting these parameters, a set of equations to solve to calculate the two scores is obtained, and making predictions for each game becomes relatively simple.”
The model we discussed earlier, referring to this year’s World Cup, works in roughly the same way: its creators assigned offensive and defensive scores to each team, eliminated the “home factor” (a parameter that increases the probability of victory for the team playing at home, which, of course, applies to all teams except Qatar), and fed the algorithm the results of numerous pre-World Cup international friendlies. Using this tool, they simulated approximately 100,000 matches, concluding that the green and gold national team will outperform the others.
Another group of researchers from the University of Innsbruck in Austria arrived at the same conclusion, albeit using a slightly different model. The insurance company Lloyd, on the other hand, used yet another model (which correctly predicted Germany’s and France’s victories in the 2014 and 2018 world championships) to predict that England will win this time (beating in the final just Brazil). Finally, the Penn group’s model crowned Belgium instead. If the algorithms do not agree…
From match analysis to scouting
But, as previously stated, predictions are only a small part of the puzzle. And not even the most succulent, at least not for those who play professional football. “According to Vanni Di Febo, football data analyst for the Italian Football Federation (FIGC, i.e. for our national football teams), science and technology enter football in at least four major areas: match analysis, scouting, accident prevention and rehabilitation, and finally all aspects of a more corporate nature.” According to Di Febo, data from cameras and GPS devices worn by players is used in match analysis to obtain variables of interest and build game patterns: Who is better at hitting headers, for example, who completes more passes, who is better at tackling, which area is played more or less? Which team is more likely to score (or concede) a goal?
“The cameras provide us with details on the position of the players and the ball thirty times per second,” explains Matteo Giacalone, Inter’s match analyst. “It is an impressive amount of data, which we process and from which we derive indicators, which we then share with the team’s technical staff.”
Our data is combined with video clips from matches and training sessions to create a video that the coach can then review and show to the team.” Not only that, but “With these data, Giacalone continues, “it is possible to construct an offensive danger index, which is essentially a linear combination of various variables – for example, the number of crosses, possession of the ball, etc. – and which measures, in a ‘objective’ way, who was the better team after the match.” “Football is a very different sport from others, for example, basketball or baseball – Di Febo continues – because it is a continuous game with low scores: just one single episode, the ball that goes ten centimeters further or further, conditioning an entire game.” Making predictions is difficult precisely because of the game’s high ‘volatility.’ Basketball, for example, is different: chance can play a role in scoring a shot, but given that many more points are scored overall, it is reasonable to expect – and indeed is – a ‘large numbers’ effect that ensures that the strongest team almost always wins.”
Big data is also extremely helpful during the scouting phase: “Using these tools, we can download and analyze the parameters of hundreds of players to determine which ones are the best to buy, both economically and in terms of technical characteristics, says Giacalone. We can narrow the search down to three or four players and then tell the coach which one is the best “. “Right now, we have the goal of monitoring all potential national team players,” Di Febo explains, “and we are scouting players from the under 15s onwards, examining all the data that comes to us from competitions: appearances, goals scored, cards.” In this way, we hope to identify the most intriguing names and encourage you to go see them for yourself.”
Then there’s the issue of injuries, and again, technology plays a key role: the data collected by the players, in fact, reveals a lot about their overall health. “For example, if we know that a fit player has an acceleration of 35 m/s2 and we see that he is unable to achieve that acceleration in training, we can speculate that something is wrong.”We can also monitor any excessive or too close efforts and notify the technical staff if the athlete is at risk of injury “. Finally, the corporate aspect, which is equally important in professional football: “We can make forecasts and assessments, for example, on player bonuses, but also on which teams are bought (or sold), the percentage of foreign players, and so on,” Giacalone explains.
