The technology of sports – how software is used in modern professional sports

There is something to be said for the beauty of professional sports. Take professional football (or soccer), for example. It is fluid in a way that becomes indiscernible if one watches it for long enough. At the elite level, this fluidity seems to be governed by a force as innate as the fluid dynamics of water. While the atomic composition of the unit remains intact, the unit dynamically adjusts its shape to yield to external pressure. Think of the dynamic triangles that in-possession teams create on the football pitch in response to opposition pressure – teammates move into space around the player in possession to create multiple passing options. If this response to game conditions seems quasi-omniscient, then perhaps it is instructive to look into the intelligence that has come to play an increasing role in orchestrating the game; and how it also plays into the fan experience. Cue sports analytics.

Sports analytics is the process of capturing data and analysing this data using mathematical models in order to deliver insights. While statistics are intrinsic to sports history, sports analytics, in its modern form, was arguably popularised in 2002 by Billy Beane and the Oakland Athletics Major League Baseball team (colloquially referred to as the A’s). Beane adopted a strategy of optimising player scouting and team selection based on rigorous statistical analysis, which came to be known as “Moneyball”. Statistical analysis revealed that on-base percentage and slugging percentage are better predictors of offensive success than the conventionally favoured qualities of batting average, stolen bases, runs batted in, running speed and defence. Because these unfavoured (statistically undervalued) qualities were cheaper to obtain on the player market than the conventionally favoured (statistically overvalued) qualities, Beane was able to cheaply assemble a team of lesser known players which took the budget-constrained A’s to the playoffs in the World Series. This objective statistical analysis of players and game plans soon became the norm. Today, data science has an influential say as to who makes the team, how the team sets up and what the game plan is in most (if not all) professional sports. The relative statistical efficiency of three-point shots and layups in basketball, for example, has meant that teams tailor their scouting, selection and game plans to making these shots.

Pre-game preparation

Real-time data analysis has also made its way onto both the training ground and the field of play. Professional athletes and coaches use software platforms from providers like ORRECO to make data-driven decisions on training and performance. Termed ‘Real-time Internal Load Monitoring’, ORRECO provides rapid biomarker testing software, which analyses (with a four- to six minute turnaround time) a pinprick blood sample extracted from the finger or earlobe. The software uses hyperoxide- and total antioxidant capacity markers as a measure of oxidative stress; and uses high sensitivity C-reactive protein (or ‘hsCRP’) as a measure of inflammation and predictor of sudden heart attack, stroke or coronary artery disease in individuals with no history of cardiovascular disease. These oxidative stress and inflammation measures provide a current view of an athlete’s physical state and response to training load.

Performing these tests at regular intervals and over a sustained period of time tends towards increased precision in the data- and artificial intelligence-driven mapping of an individual athlete’s oxidative stress and inflammation ranges and helps to identify their thresholds. Results are delivered through ORRECO’s Zone software platform, which provides data visualisation and actionable Insights in the form of instant alerts for any increased risk of injury, illness or strain as well as advice and recovery strategies. ORRECO’s evidence-based bio-analytics software offerings are informed by enduring peer-reviewed research and are used across different sporting codes – North America’s MLB (Major League Baseball), NHL (National Hockey League), NFL (National Football League) and NBA (National Basketball Association); the English Premier League (football); the Olympic Games (multi-sport); and the ATP tour (tennis).

In sports such as team pursuit track cycling, where optimising efficiency and power output is the name of the game, real-time data analytics empowers coaches to deliver insights directly to cyclists while they are on the training track. In an example that leveraged the power of data analytics, the internet of things (IoT) and mobile communication, the USA Women’s Cycling team partnered with IBM and Solos™ smart eyewear to prepare for the 2016 Summer Olympic Games in Rio de Janeiro. The solution was predicated on the transmission of data from a mobile app on the rider’s person using IBM’s cloud-based Watson IoT Platform, the IBM MobileFirst Platform and IBM Analytics. Key metrics and signals sent from coaching staff were displayed to the rider in real time on the heads-up display of the Solos™ smart eyewear – enabling the rider to adjust output on the spot with little to no disturbance in riding flow. The cloud based technology allowed real-time analysis and communication irrespective of training location; as well as remote engagement when the team was geographically separated

Match day

An example of sports analytics at play on match day is the use of Opta data (a subsidiary of the Stats Perform sports data and analytics company) in prominent football leagues around the world, including the English Premier League and La Liga in Spain. Opta uses trained match data analysts to collect detailed data on up to 3 000 actions per match using bespoke Stats Perform data collection technology. Every time a player touches and controls the ball, data points are captured on the player’s identity, what they did on the field, where they did it and how they did it. This data is captured in real-time through a synergy of human annotation, computer vision and AI modelling. The resultant Opta data feeds are distributed in real-time to a global clientele including professional sports teams and organisations, broadcasters, media and betting industries.

Feeds are filtered to different detail levels depending on the user’s requirements. Opta’s core level data feeds deliver result and fixture information. Its classic level feeds provide individual player statistics tailored to powering a fantasy game. And its performance level feeds track thousands of individual data points (including full x, y location coordinates for every on-ball event) that enable interactive displays typically seen in expert post-match analysis. These feeds also enable multi-channel audience engagement and inform pre-game result- and individual performance predictions. They also offer richer in-game and post-match storytelling.

Engaging the 12^th man

Opta Predictions uses AI predictive models to analyse historical, in-season and in-game data. Models are constantly recalibrating in response to all kinds of data inputs, including:

Team-specific data – historical and recent head-to-head performance, team strengths and weaknesses
Player-specific data – historical and recent matchup performance, player strengths and weaknesses
Game location and conditions – home ground advantage, time of day, time of season
Game situation – current score, remaining time, regular time versus extra time

These predictive models help to fuel pre-match debates on team selection and relative standing of the opposing teams. They also provide live win probability that responds in real time to actions taken in the game. This enables broadcasters and publishers to craft informed narratives to draw the audience in.

Using Opta, broadcasters and publishers can alert fans to highlight moments in each match using ‘Qwinn ratings’. The purpose of this is to enhance content discovery – drawing fans to tune in to broadcasts and streams. Match events are ordered by real-time personalised fan excitement ratings, which are determined by a number of factors – the fan’s team preference, player milestones, total match score and lead changes.

Calling the game

Software has enabled a variation of the Video Assistant Referee (VAR) in many sports codes. Cricket has long had the third umpire whose job is assisted by video replays and software such as Hawkeye ball tracking. Tennis has also adopted Hawkeye challenges for marginal calls. Football has been somewhat of a laggard in this area – with VAR having only been introduced in the English Premier League in the 2019/20 season. The aim is to reduce the number of errors in key match decisions by allowing intervention (by the VAR) to the on-field adjudication of a match when there is a “clear and obvious error” or a “serious missed incident” in the following match-altering situations: goals; penalty decisions; direct red-card incidents; and mistaken identity. In the season in which VAR use was introduced, the accuracy of key match decisions rose to 94 percent – an improvement on the 82 percent accuracy recorded in the preceding season. The VAR uses both real-time and slow-motion replay technology and multiple camera angles in order to make an objective decision that they then convey to the on-field referee.

Software use is inextricable from modern professional sport. Data and sports analytics have come to define the way in which athletes and teams compete and, thus, what those athletes and teams look like. The efficiency, intelligence and sheer reach of this software has meant that it has come to augment the professional sports experience in a way that is so natural that it is almost imperceptible.