Build or Buy: Assessing Efficiency of Individual Clubs Developing Own AI-powered Injury Forecasting Models

While Sevilla's model of building out data informed technology solutions has had success to-date, the club, as well as any other individual club, is likely to run into a few critical issues when it comes to building out an AI-driven solution to address injury risk forecasting and workload management.

Sevilla are a forward-thinking club.

Having made data a critical piece of their decision making processes, the club has implemented an ambitious innovation policy across their performance and commercial departments.  

This is best exemplified with the successful development of Transfer Tracker. The tool helps track player transfers to help clubs to recoup millions of euros worth of secondary transaction fees. The system, which was initially developed and used in-house by Sevilla FC’s data and legal departments, was brought to market alongside LaLiga Tech in December. 

Last year also saw the LaLiga side agree a unique five-year partnership with FC Bengaluru United, looking to tap into the wealth of IT talent within the Indian city.

However, the key point of interest for data’s impact has now turned to injury prevention.

El Mister recently reported that injuries have cost Sevilla FC, on average, €30m ($34m) per season, from a combination of salaries lost and associated medical expenses.

Club president Jose Maria Cruz is quoted on the issue: “If we manage to reduce injuries by 10 or 15%, imagine the savings we will have.”

Whilst the Spanish club’s model of building out data informed technology solutions has had success to-date, the club, as well as any other individual club, is likely to run into a few critical issues when it comes to building out an artificial intelligence (AI)-driven solution to address injury risk forecasting and workload management. 

We’ve identified three core hurdles to this approach; data volume, data quality, and talent resource.

  1. The Data Volume Requirements are Enormous.

As Cruz was quick to add himself, “It is relevant for the business of the teams, but we have to do it with data from all possible teams, our metrics are not enough.”

Therein he identifies the bottleneck that has an enormous impact on whether AI can reach its potential – – an understanding of the fundamental inputs AI needs in order to provide the requisite usable outputs. 

Typically, many thousands of ‘man years’ of player data are required for an algorithm to become accurate enough to be usable by a club on a day-to-day basis. It is not feasible for individual clubs to meet the data volume requirements to build or train an injury forecasting algorithm.

Gavin Benjafield, Performance Director at LAFC, has been working directly with Zone7 and explains this best; 

“We [LAFC] are in our fifth season as a professional team, but even a club with 100 years of history cannot match the multiple data points from multiple clubs and environments that Zone7’s AI platform offers,” he said. 

“When we were ready to determine whether to adopt Zone7, that context made the decision relatively easy for us.”

Gavin Benjafield (Performance Director, LAFC)

Zone7 has collected hundreds of millions of hours of human performance data and tens of thousands of injuries collected in partnership with many of the best football teams in the world. 

The sheer volume of data required for Sevilla FC building out their own alternative would mean establishing a direct collaborative process with other clubs – some of which would likely be rivals. 

The reality in sports today is that a massive dataset, with the data connected (different devices, etc.) is not available for teams to use in-house. Zone7 has been able to achieve this as a trusted partner.

  1. Not All Data is “Machine Learning-Ready” 

Datasets used for AI need to be validated for quality and enhanced to ensure any outputs generated are contextual, meaningful and potentially impactful. The ‘Garbage In, Garbage Out’ (GIGO) maxim is at the forefront of any data science quality control process. 

The data needs to be collated, cleaned, and standardized across the many different data sets from different vendors. Processes like checking for value spikes, being consistent with thresholds/bands, and managing the athletes in the correct sessions and drills. 

It is not uncommon for Zone7’s automated “data diagnostics” to uncover glitches that have a negative impact on the data integrity in a specific environment. More so, different teams use different devices or collection protocols, which requires a “stitching” process to take place for the full potential of the dataset to be available. 

Zone7 has built out the tools and skill sets to assist practitioners and their organizations to ensure data quality is sufficient and has put measures in place to ensure each new contributing dataset is not only an appropriately valid dataset, but also one where ad-hoc anomalies are identified and removed so that they do not pollute the data lake. 

For an individual club to take on this responsibility, there is an incredible amount of time investment required. Increasingly, teams seeking to conduct their own data science projects are turning to Zone7’s audit and elite services to assist them on their own, independent journey’s.  

  1. A talent and capital intensive process 

For the algorithms to improve, a club would also need to “feature engineer” new variables that add context to the data, create data anomaly protocols to limit the impact of faulty wearables, and have a team to supervise the models for continued improvement. Also, “operationalizing” machine learning in a live software environment that runs smoothly every day requires significant investment. 

The team required would need senior data science and machine learning experts as well as a team of engineers. The wages for these professionals might cost as much or more than the players on the team. The ROI does not tie out. 

Sevilla is aiming to address this issue through their partnership with FC Bengaluru United, which is based in the ‘Tech Capital of India’, and through their joint Hackathons with the club. However, this is still external talent working on the problem for only a few days of the year.

By using Zone7, teams are tapping into a resource that is powered by 30+ full-time experts within this field. To build out that sort of capability in-house would be extremely difficult to do, not to mention the significant expense of it, too.


Sevilla are, indeed, a forward-thinking club, with the innovations they are looking to establish in-house. But being forward-thinking also means recognising when things can be done more efficiently and effectively by utilizing external help. The cost of hiring just one team member that a club would need to build out an in-house solution like Sevilla’s can be outsourced to Zone7. 

Zone7 offers an AI-solution that is pre-built, but completely customized for the end user. The company has a global team of world class sports scientists, data scientists, and machine learning experts who have built the hard part already. To bring the AI to life in a sports team just requires a few weeks of tuning, rather than years of building. 

We are ready to help. If you are interested in improving player availability and performance, get in touch today.

Ready to Elevate your Performance? We provide a comprehensive suite of products and services utilizing data and AI to enable greater performance and durability amongst high-performance teams.

To find out more, Book a Demo today.