A short history of data analysis in football
How data, the Internet of Things, and analytics are helping to improve the performance of individual players and football teams
Sponsored Feature Sports and technology are increasingly intertwined. A combination of wearable devices, high-definition video tracking applications and manual input are collecting large volumes of data about player movement and performance, which can be used to improve individual and team results.
But it takes significant IT processing muscle and data science expertise to analyse all that information, combine it into a meaningful whole and present it in a way that delivers critical insight, potentially in near real-time.
In the first in a series of five, this article shows how data analysis has been used in football, from the earliest days of international teams and top club sides tracking passes, shots, crosses and corners, to modern techniques that see players using wearable devices to measure their performance, workload, fitness and fatigue.
Inside the Brazilian mind
One of the earliest examples of data science being used to measure potential football performance were the psychometric tests used by Brazil's team management ahead of the 1958 Fifa World Cup in Sweden. At that point, this type of approach was still in its infancy and the scientific conclusions reached were never fool-proof. Tests recommended that arguably the world's greatest-ever player, Pele, should be omitted from the Brazil squad for example, having been deemed "infantile" and "lacking fighting spirit". Thankfully the team manager selected Pele regardless and he went on to score six goals during the tournament, including two in the final when Brazil won the trophy.
It's important to remember that football matches can be won or lost on the tiniest of margins however, and subsequent applications of data science proved more successful.
The AC Milan 'Mind Room' unit in Italy became legendary after it was set up in the late 1980s, combining stress relief therapy with cognitive training and neuroscience driven by player data. Its reputation was certainly helped by a string of AC Milan championship and cup wins throughout the decade. Some players publicly praised the impact the unit had in helping them turn "positive thinking" into a winning mentality under pressure.
IoT improves data gathering and analysis
The use of data science in sport has been boosted by technology systems which make it quicker and easier to capture a player's performance metrics and better interpret and share results. A prime example involves Google Cloud working with Technical Directorate staff for England's Mens and Womens international football teams at The FA, a collaboration that helps secure access to databases, processes, functions and compute resources which combine to analyse large volumes of data pulled from different sources.
A separate partnership also sees The FA working with IT services firm Cognizant on several other digital transformation initiatives designed to support improved data collection and analysis.
David Ingham, head of media, entertainment and sport at Cognizant, says: "When considering the use of data within football, it is crucial to think about it in two parts - data gathering and data analysis. There is no doubt that data gathering in football has improved over the last few years, largely thanks to the increase in IoT (Internet of Things) technologies.
"From the introduction of wearables that track players' movements during games and practice, to sensors in the ball, and all around the stadium, there is now a plethora of data sources constantly feeding us information from all sides."
Many teams have invested heavily in automation technologies which ensure that all of this data can be collected and aggregated with speed. The use of computer vision, which allows coaches to request to see all shots or all passes by a certain player on-demand and in a matter of seconds, is now common. This reduces the time previously spent manually reviewing old footage and means coaches and managers can be more efficient when developing training plans or match tactics.
Cognizant has been helping teams design and implement these processes alongside its partners in the football ecosystem, to try and equip teams at the grass-roots level of football with the tools they need to get the same level of insights as the larger teams.
However, while teams have access to their own data, they often lack the same level of insight into their competitors, hampering the ability to effectively prepare for future matches, says Ingham. To achieve this, he says, football could take inspiration from the likes of F1, in which certain telemetry data is available to all teams, which in turn helps to create a more level playing field.
Not having this level of access in football makes it harder to, for example, run computer simulations on how competitors might react during a game to substitutions, or other tactical changes. This means that despite there being huge volumes of information out there, many coaches without access to it are still relying on "gut feeling" when making decisions during a game, when they could instead be listening to what the data tells them.
"I believe this will be the next stage of development for football data analytics, and it most likely won't be long before clubs are hiring data scientists at the same level as they now hire coaches and physios," Ingham says.
AI and automation on point
Ryan Beal is CEO at SentientSports, a company which focuses on AI (artificial intelligence) decision-making services in football and other sports.
He points out that many clubs don't have the expertise or processing power to deal with billions of data points per season that are collected. So new AI-driven services may have to be used to help process this data for clubs, especially at lower levels where they do not employ specialised data scientists.
This includes assessing whether certain players are going to be suitable for the teams wanting to sign them.
He says "The type and breadth of data has evolved significantly in recent years, moving on from 'box-score' stats that you might see on TV broadcasts to event-by-event data - around 1,500 events per game - and to now full player tracking using computer vision and GPS technologies, that can collect player positions 20 times a second across a game or training session to extract more physical metrics."
Another firm, Spiideo, provides clubs like Inter Milan, AS Roma and Brentford with video performance analytics software. It also supports a range of men's and women's international teams, including African champions Senegal.
"We see AI continue to prove its efficacy," says Fredrik Ademar, vice president of product at Spiideo. "Predictive analysis is already beginning to appear, with AI able to calculate probabilities and assess athlete performance in real-time. We've believed from the start that video and data needed to be live and freely accessible. With greater integration of video with AI data, these systems are able to monitor and alert coaches of tactical and performance advantages."
Ademar says that with the help of automation, video analysis is enabling all teams to employ and analyse their own performance. "Women's football in particular has been making great strides by adopting video-based analysis. Sweden's national women's team uses our system to enable coaches, analysts and players to review performance during and after matches."
Inter Milan at the edge
Inter Milan also uses GPS and video tracking tool to collect vast amounts of player data around speed, acceleration, power, distance and other areas, within a granularity of one millisecond.
The installation of edge servers has increased the speed of data processing. The data is packaged into detailed reports that are provided to coaches, physical trainers and technical analysts, who then use them to identify tactics that could be most effective when facing specific opponents.
"We can now track every moment of every game, every training session, every player and process," says Lorenzo Antognoli, director and head of technology at Inter Milan. "The idea is to analyse the physical performance of each player according to the different technical and tactical situations on the pitch. All the data is then ingested, enabling us to store and process it. Then we apply algorithms to identify patterns and provide insights through dynamic reports."
The next step, he says, which is expected to happen soon, is to use the edge technology so the club can process data in real-time during a match.
From modest beginnings, it is clear that data, and the technology behind it, is now key for the future of football performance, and there is still a lot more to come.
Sponsored by Google Cloud.