2021 Volume 24 Issue 1

Editorial: How AI can Help Reduce Sports Injuries

Editorial: How AI can Help Reduce Sports Injuries

By most standards, the sporting industry is big business with annual global revenues approaching $500 billion. Professional team sports accounted for approximately 20 percent of that total in 2019 with European soccer leading the way with a 30 percent share followed by the National Football League (NFL) with 15 percent. To that end, fan interest in professional sports has continued to grow over the past decade. Alas, the rise of COVID-19 at the beginning of 2020 has completely altered this landscape. Few industries have been hit harder by the pandemic than professional sports. For example, estimates suggest that NFL revenues could drop by 25 percent because of teams playing in empty stadiums. This figure could grow even higher due to the current political unrest, which is impacting TV viewership. Furthermore, under current contract arrangements involving many professional leagues; lower revenues would reduce the salary caps, which could have a significant impact on player employment opportunities and, in turn, viewer interest.

Although attendance and revenues continue to decline, the challenges associated with player injuries continue unabated. Not surprisingly, the NFL heads the list of injuries on a per game basis. This outcome is due to the nature of the sport and the number of players engaged. Two of the most common football injuries are anterior cruciate ligament (ACL) tears and concussions. Professional sports spent over one billion dollars on player injuries in 2019. It is hard to imagine the costs associated with player injuries declining any time soon under the current game protocols and the size of the teams needed to play the game. Indeed, the ability of teams and players to obtain injury insurance is becoming increasingly difficult, which naturally simply drives up the costs.

One could argue that given the overall present economic instability the ownership and player associations could agree to reduce the size of the teams as a vehicle for cutting costs. However, that would simply expose the remaining team members to a higher-level risk of injury. So, within this context, is there a better way of reducing the incidence of player injuries and at the same time maintaining the integrity of the game? Artificial intelligence (AI) offers one promising approach. The task at hand in reducing player injuries is somewhat like predicting customer churn, which has been extensively studied using artificial intelligence-based machine learning algorithms. The central idea behind churn analysis is to identify customers that may be switching to another supplier and to enact specific strategies to ameliorate this phenomenon. In team sports the idea is to identify a player that could be a candidate for injury and take appropriate operational steps (e.g., reduce playing time and alter the player’s training regimen). Additionally, if a player has become injured, the same modeling process can be used to specify the optimal recovery method based on the best long-term interests of the player.

It’s not whether you get knocked down; it’s whether you get up. – Vince Lombardi.

Machine learning, a subset of AI, is the science of discovering and communicating meaningful patterns in data and supporting the development of actionable plans. Unlike AI, most machine learning schemes involve some human involvement. Specific human tasks include preparation of the database, selection of the appropriate algorithms, actionable knowledge discovery, and interpretation and implementation of the results. Today, machine learning is experiencing increased use throughout professional sports, including outcome predictions, play selection, yield management, and injury risk assessment.

Yield management, which has its origins in the airline industry dating back to the 1980s, focuses on the notion of maximizing the use of existing capacity through variable pricing. As applied to sporting events the idea is to engage in dynamic seat pricing based on a variety of conditions, such as the day of the week and strength of the opponent. The common denominator associated with each of these sports analytics applications is a vibrant and extensive database. The database needs to be characterized in terms of a large number of candidate predictor factors. The term candidate is used since not every factor identified will end up in the final model. With respect to predicting sports injuries, the relevant factors can be grouped into three broad categories: Environmental, Player, and Organization. Examples of environmental factors include, but are not limited to, game weather conditions, rules, time in the season, and publicity. Standard player variables are age, tenure with the team, and prior injury history. Some team organizational elements include winning percentage at time of injury, coaches’ tenure, and nature of the training process.

Typically, the most straightforward approach is to specify which players are at risk of injury based on the factors outlined above. However, an even more interesting approach is to identify the type of injury(s) that may occur to a given player. This methodology would allow for more precise targeting of amelioration actions pre-injury. Of course, this whole approach depends heavily on developing a consistent definition of what constitutes an injury. Presently, most sports injury reporting systems lack a consistent conceptual structure, especially between the various professional sports leagues. For example, Major League Baseball (MLB) lists players out for a minimum of seven days, while the National Hockey League (NHL) typically only reports upper or lower body player injuries. As a result of the COVID-19 pandemic, the NFL adopted special rules for the 2020 season that allow a team to remove any player with a football or non-football injury from the roster for three weeks after which they will be eligible to return to practice. One of the goals of this policy is to ensure the availability of enough players in the event of subsequent waves of the virus.

It ain’t about how hard you can hit. It’s about how hard you can get hit and keep moving forward. – Sylvester Stallone (aka Rocky Balboa).

The use of machine learning as applied to the ongoing challenges of reducing player injuries can be characterized by the following steps: 1) Identifying players with a high degree of certainty of being injured in the future, 2) Specifying actions to minimize the incidence of such injuries, and 3) Developing cost-effective amelioration strategies once a player has been injured. One of the shortcomings associated with many of the machine learning algorithms is that they have the “appearance” of being a black box, which makes understanding the results somewhat problematic. For example, in the financial lending sector, some otherwise well-performing algorithms are not used in the credit assessment process since the lender cannot effectively explain to the applicant why credit was denied.

Decision trees offer some relief in this regard since they provide a visual rendering of the decision-making process and the relationship between the variables. Trees are formed by a collection of rules based on values of the candidate variables in the modeling process (e.g., player’s age). Rules are selected according to how well splits based on variables’ values can differentiate observations of the target variable, which in this case is player injury. Once a rule is selected and splits a node into two, the same logic is applied to each dependent node.

To illustrate this process, consider the following NFL situation. Suppose the machine learning algorithm has identified player injury history, game weather conditions, and player position in that order of importance. Then the algorithm could be used to predict the chances of a specific offensive lineman, with an injury history being injured during a game marked by inclement weather conditions. The coaching staff could use this prediction as a basis for making an appropriate substitution. The same information could be employed for predicting how long a player that was identified as being at significant risk of injury would be on the disabled list, which, among other things, could impact the outcomes of future games.

To underscore how AI is catching on in the professional sports world, in December 2019 Amazon and the NFL announced an arrangement wherein NFL player health data would be analyzed using Amazon’s Cloud Unit artificial intelligence and machine learning technology. The basic objective is to develop a wide array of medical and health cost-effective solutions for the players. Hopefully, the application of artificial intelligence in this matter will not only contribute to improving player safety and team performance but will also provide the leadership with much-needed expense relief during these troubling times. These same injury prevention and amelioration technologies can be applied to a wide range of sporting activities starting from K-12 through college, which will allow the participants to better appreciate the true value of sports, namely the pursuit of excellence and how to handle adversity.

No athlete is truly tested until they’ve stared an injury in the face and came out on the other side stronger than ever. – Anonymous.

 

This article is dedicated to the memory of Dr. Thomas J. Dudley, a co-founder of the Graziadio Business School and an avid sportsman!

Print Friendly, PDF & Email
Author of the article
Owen P. Hall, Jr., PE, PhD

Owen P. Hall, Jr., P.E., Ph.D. is a former Corwin D. Denney Academic Chair and is a Professor of Decision Sciences at Pepperdine University’s Graziadio School of Business. He is a Julian Virtue Professor and a Rothschild Applied Research Fellow. Dr. Hall received the Harriet and Charles Luckman Distinguished Teaching Fellow in 1993, the Howard A. White Teaching Excellence Award in 2009 and 2017, and the Sloan-C Effective Teaching Practice Award in 2013. He is the vice-chair of the INFORMS University Analytics Programs Committee. Dr. Hall has more than 35 years of academic and industry experience in mobile learning technologies and business analytics.

More articles from 2021 Volume 24 Issue 1

After the Covid-19 Economic Crisis

Introduction Value for Money (VfM) analysis helps governments decide whether it is more cost-effective to do a project through traditional procurement, or through PPPs. State government financial departments, project teams, and senior management can run this type of simplified VfM analysis with the help of a few consultants in order to select the ideal type … Continued

Related Articles
After the Covid-19 Economic Crisis

After the Covid-19 Economic Crisis

Introduction Value for Money (VfM) analysis helps governments decide whether it is more cost-effective to do a project through traditional procurement, or through PPPs. State government financial departments, project teams, and senior management can run this type of simplified VfM analysis with the help of a few consultants in order to select the ideal type … Continued