“You’re configured to referee!”

Ever since the first match of association football was played in 1872, the adage that referees are incompetent and ruin the spectacle has pervaded. The furore around inept refereeing decisions has forever punctuated football discourse.

For years, subjective split-second decisions have been scrutinised to the Nth degree, as fans can’t possibly perceive the idea that a referee could not see the slight deflection off the defenders’ hand, through a scrum of jostling bodies littering the penalty box.

By the 2010s, the situation had reached a critical point. As other sports embraced technology, football continued to inhibit refereeing performances by refusing to implement video technology to assist the decision-making process. At the same time, the clubs who were performing the best on the pitch had embraced advanced technology to inform recruitment and performance analysis, maximising the value of their decision making.

Belatedly, football’s law-making body IFAB permitted the use of Video Assistant Referee (VAR) technology in 2016, to widespread celebration and the expectation that, finally, the standard of officiating would improve – and the omnipresent controversy surrounding referee decision-making would cease to exist.

By finally embracing technology, the individuals responsible for the governance of the game were recognising that technology and data, when utilised correctly, can add significant value to performance and output. 

Fast forward to 2023, seven years after IFAB originally approved the use of VAR technology, and the problems are far from solved. In fact, much like individual refereeing decisions, the issues of VAR have been further magnified. Far from being the solution to all officiating ills, the introduction and implementation of VAR has been deemed as a failure. Fans, pundits, players and managers have all come to the realisation that, while the use of technology improves visibility, it is still humans making decisions and, ultimately, humans are flawed. 

So, what now? Where does the game go from here? The ongoing issues that have rolled into the 2023-24 season have even seen calls for VAR to be scrapped. That the conversation is happening is indicative of an implementation approach that has failed. 

A solution is needed, and it will come in algorithmic form. 

IFAB, Fifa and other governing bodies may not need to look far for the solution to the age-old problem of officiating. There are 151 years of evidence that when it comes to officiating, human intelligence is insufficient. Could the answer now more obvious than ever? Just as it has been mooted as the answer to complex societal problems, artificial intelligence is being positioned as the tool to save the day.

Performance Analytics 

The data revolution that has taken place within professional football has been gathering pace since the formation of the Premier League in the early 1990s. For the first time, clubs were able to use video technology to track core statistics, such as match possession, aerial duel success percentages and passing accuracy percentages. 

The game really changed in 1996 when Opta Consulting began to manually collect data from live matches and retrospective video footage to exponentially grow the level, scope and extent of available datapoints, which were then provided to the clubs and the media. This had an almost immediate impact on many decision-making processes at clubs. Managers were making transfer decision based on data, rather than  just their own eyes. Players were tracking their own performance and fans became obsessed with pass completion rates.

Fast forward to 2023 and football and data have become synonymous. Fans in pubs and cafes around the world now discuss xG and xG per90. Social media has become awash with amateur analysts creating detailed infographics illustrating underlying performance, trends and future stars. 

Performance analysts have access to vast databases, such as Statsbomb, and video catalogues, such as Hudl, which has resulted in the data team becoming a key element of value creation from a performance perspective. Every professional club now employs its own team of analysts, responsible for processing and analysing data to improve decision making around recruitment, opposition analysis, injury prevention and a plethora of other niche performance scenarios. Incremental value creation is seen as the promised land for professional football teams. Player analytics has become mainstream, with broadcasters and prominent publications featuring advanced statistics data, visuals and infographics in their content, powered by media data rights packages that are agreed with specialist sports data providers.

 The use of advanced descriptive statistics has become ubiquitous since xG emerged to measure the performance of current players. Historic data is used to attribute a value for every shot and certain passing actions and the likelihood that a goal will be scored from that specific position on a pitch.  

The extent to which descriptive analytics can measure the underlying performance of players and teams has continued to develop, with supplementary statistics such as xGA, xGChain and XGBuildup now utilised to assess the non-goalscoring actions of players and to understand the value they add to the overall attacking build-up phase. 

The development of descriptive analytics has been fuelled by specialist sports data providers who make a significant proportion of their data publicly available. This has contributed to the increasing use of advanced descriptive analytics in football-related discourse. 

Deep Learning Analytics 

While the emergence of xG was seen as a new dawn for data analytics when it first emerged in 2012, just over a decade later the sport is on the cusp on the next evolution in data analytics. The approval of wearable technology by Fifa in summer 2023 will facilitate the proliferation of deep-learning models. 

Deep-learning models are already present within football analytics. The astrophysicists and Ivy League graduates that make up the data teams at clubs with the most sophisticated analytics departments have been using deep-learning models to value each action a player makes during a match, to assess the value that they add towards the team’s performance objectives. 

However much of this has, so far, been through the use of AI-powered models based on secondary data collected by sports data providers and gathered from analysis of broadcast content. Deep-learning models utilise this data to run simulations which assess the value that a player has on the outcome of a season, or overall team performance based on the value of their performative actions. One such example is passing ability, which is measured according to the impact that each of an individual’s passes has on the chances that the team will score, or concede. Using deep reinforcement learning models to value every pass, analysts are able to apply this data to context-driven algorithms, such as age, physical attributes and competitiveness of the match scenario, to derive actionable analytical outputs which inform decision making. 

While this level of analytics has been taking place for a number of years, the decision by Fifa to allow players to wear tracking technology on their boots will augment the deep-learning capabilities of analysts even further. As wearable technology advances, it is inevitable that the data available to analysts will become increasingly advanced, to a point where every touch, pass and movement will be scrutinised to the Nth degree. A player’s level of technique will be able to be measured as fact, with a player who uses 60 per cent of his instep when passing the ball encouraged to work to reduce that figure to 57 per cent in order to reach the optimum passing technique. 

The real data revolution will occur as wearable technology advances, and deep reinforcement learning analytics will be used across professional teams and academies globally, to work on a players’ technique to minimise the chances of failure and reduce uncertainty to a point where only players who score within the 95th percentile of shot velocity, passing technique variety or tackling veracity will make it.    

Fast forward to 2023 and football and data have become synonymous

Algorithmic Referees

So, what does this mean for referees? While this data revolution has been occurring in player performance, referees are only just now using video technology. The same referees who have been trained to use naked-eye decision making for their whole career are now told to watch the action on a screen and use the naked eye to make the same subjective decisions. 

While data has proliferated across professional football performance analysis, referees have been left in the cold, still expected to make split-second decisions using only their eyes and a set of five camera angles. 

While the neglect that the refereeing profession has had towards data up until this point is barely comprehensible, the nadir is close. 

With Fifa’s decision to allow AI-driven wearable technology to be used in official matches by players, the white knight for referees creeps ever closer. 

With wearable technology comes real-time data and with AI-powered wearable technology comes advanced Big Data. This will track every movement a player makes and provide instantaneous feedback to a centralised data lake and an algorithmic ‘referee’.

Real-time data will track a player’s movement to the millimetre. It will geolocate every player on the pitch to the exact piece of grass that their boots are placed on at any given moment. And it will assess the timing, velocity, volume and veracity of each touch of the ball. 

The future of refereeing will be transformed. In the short to medium term, progressive use of data and technology will enter the game to aid the decision-making process. Referees will be drip-fed additional data and technology which will solve specific decision-making processes around specific incidents, removing ambiguity in a way that goal-line technology solved that specific issue instantaneously. Referees will still be adjudicating, and the skillset required to officiate will remain broadly the same, while those in the VAR room become, incrementally, more data-orientated. As the availability of real-time data continues to grow, the art of refereeing will become simultaneously both more complex and easier, as more information clouds the decision-making process but produces more accurate outcomes.

Eventually, in the long term, the incremental increase in data-driven officiating will result in the reduction of referees to a role more akin to ‘umpire’, enacting the decisions which are made by the data and the algorithmic referees. As officials continue to utilise the technology, the skillset required to be a referee will gradually shift from real-time adjudication on player actions, to the ability to interpret data and act accordingly. 

Contentious offsides will be a thing of the past, as every player will be geolocated on the pitch at the exact moment the ball was passed. 

Fouls and cautions will be able to be judged objectively, according to the velocity of physical impact between two players. 

Throw-ins will be automated, as the speculative claim that the opponent got the last touch will be immediately overruled by the omniscient algorithmic referee. 

Forget five-minute VAR reviews, as the algorithmic referee will be able to attribute a value on every player’s action instantaneously, eradicating ambiguity. 

It could even reach a point where deep-reinforcement learning will allow referees to predict when a foul will be made. 

The days of subjective decision making will be over, the officiating accuracy that the game has always strived for will be obtained and contentious decision making will be a thing of the past. 

In the not-so-distant future, as spectators sit there, the sound of the AI-powered algorithmic referee’s decision explanation ringing around the stadium tannoy. Memories of the halcyon days of pub debates over refereeing decisions, evocative reactions to a throw-in given the wrong way and the split-second glance at the assistant referee’s offside flag after a last-minute winner, will come flooding back and nostalgia will set in. Was it really that bad before?