STATSports
Nominated Award: Best Use of AI in Sector
Website of Company: https://statsports.com/
STATSports has grown from its humble origins in Ireland to be the world-leading provider of global navigational satellite systems (GNSS) for player tracking and analysis equipment. With over 12 years’ experience, working with individual clients and elite teams across different sports, we now have over 120 employees spread around the world, with offices in Ireland, London and Chicago.
STATSports system is a customer driven platform, composed by individual wearable devices (APEX Pods) and proprietary software (Pro Series – Sonra for teams and Athlete Series for individual consumers). The unique, user-friendly interface was designed and shaped alongside some of the world’s top strength coaches, medics and sports coaches. It is the only system to offer sports specific software for football, American football and rugby.
Our Pods have been tried and tested by many of the top teams and research facilities the world over and time and again results prove our system to be the most accurate, reliable and consistent device on the market.
The Pods contain the latest GNSS technology available, alongside inertial measurement sensors (Accelerometer, Gyroscope and Magnetometer). By combining the most advanced GNSS technologies with powerful inertial measurement sensors, STATSports delivers more than 260 metrics with high precision, both live and post session. It is the only system in the market to assure a perfect correlation between live and downloaded data, ensuring accurate data is delivered at key times to make informed decisions.
The STATSports’ system is used by over tens of thousands individual consumers as well as 600 teams worldwide, including the Irish Rugby Football Union (IRFU), Football Association of Ireland (FAI), Brazil FA, England FA, Portugal, FA, Juventus, PSG, Liverpool, Manchester City, Manchester United, USA Rugby, New York Yankees, New York Jets, Miami Dolphins and the Board of Control for Cricket in India.
Reason for Nomination:
In recent years interest in sports analytics has increased dramatically, reflected in both the number of research papers published in this area, and in the increasing number of annual sports analytics conferences. Since FIFA approved the use of GNSS tracking devices in matches in 2015, spatial-temporal data has become one of the most demanded commodities in sports analytics. Nowadays, GNSS technology is used extensively in elite sporting organizations across a multitude of sports, therefore the availability and use of spatial-temporal data has increased exponentially, together with other match data like video and event data.
One of the main players in GNSS technology is STATSports. Our proprietary APEX pod, comprises of a GNSS antenna and Inertial Measurement Units (IMUs) including: a tri-axial
accelerometer, a tri-axial gyroscope and a tri-axial magnetometer. High quality GNSS and IMUs data in combination with cutting edge analytical techniques, namely Artificial Intelligence (AI) – Machine Learning (ML) allow a deeper understanding of different aspects in team sports.
To date, the majority of research in the area of statistics and ML using team sports data is related to match result prediction or overall team performance, leading to a lack of useful and usable research using GNSS and IMUs data in team sports and hampering the coaches understanding and adoption of this type of data. This is well documented – and
one of the biggest barriers to sport science and GNSS technology in different team sports is obtaining the coach buy-in to the data that is collected.
STATSports is using AI in its Pro Series products to help overcome these limitations, by providing innovative insights on the use of GNSS and IMUs data that was previously
overwhelming for the coaching staff. STATSports have developed an innovative approach to analyse player’s GNSS and IMUs data, mainly focusing on the behaviour and workload of each athlete, individually and collectively. This approach can be divided in four main aspects:
1) Output of new performance metrics 2) Prediction of training load
3) Automatic classification and labelling of individual sporting actions
4) Automatic detection and classification of teams’ formation
The complexity and the type of data needed for each of the different sports performance analysis aspects vary, and so does the AI solutions used for each one. Due to the need of high computational capacity and the enormous amount of data, these solutions are implemented in a cloud environment.
The use of a cloud environment opens new perspectives for the individual consumer side of the business – Athlete Series. STATSports will be able to provide state-of-the-art scientific research and the most advanced AI solutions to its individual customers based on their own data. Simply, individual customers will have access to the same tools as professional athletes adapted to their own performance and needs.
STATSports’ pioneering use of AI in different aspects of team sports performance is raising the performance level of professional and non-professional athletes. For this reason, I truly believe STATSports is a strong contender to the Best Use of AI in the Sports sector.
Additional Information:
The complexity and the type of data needed for of each of the sports performance analysis aspects vary, and so does the AI methods used for each one. For a better understanding of each of them, a brief overview and some examples where STATSports i using AI are given:
1) Output of new performance metrics – The accuracy of GNSS is degraded indoors, however IMUs have made positioning calculations possible in these venues. STATSports is using Machine Learning – Neural Networks to calculate position using only spatial-temporal IMU’s data. Using neural networks, implemented in Microsoft Azure (due to the need of high computational capacity and the enormous amount of data), we are able to accurately estimate players distances and velocities in indoor environments, extending the application of our devices to different sports (e.g. Basketball, Handball, Futsal, etc.) and different sporting venues.
2) Prediction of training load – Managing an athlete’s training load appropriately is crucial to obtain a positive training adaptation while reducing the likelihood of sustaining an injury. STATSports is using different Supervised Machine Learning Regression and Classification methods on spatial-temporal GNSS and IMUs data to predict different teams’ training load of each training session. By allowing the end users to select a minimal amount of information: Session Type; Drill Category; Drill Title; Duration and Players involved, we are able to predict training load metrics like: Total Distance, Distance per Minute, Maximum Speed, total number of Sprints, total number of Accelerations and Decelerations, among others.
3) Automatic classification and labelling of individual sporting actions – Using different Supervised Machine Learning Classification methods on spatial-temporal GNSS and IMUs
data allowed STATSports to develop algorithms that automatically detect and classify specific sporting events like:
a) Different cricket bowling deliveries – Fast and spin bowls.
b) Different Goalkeepers actions – Jump, Dive High/Low, Right/Left , etc.
By classifying different high intensity actions we aid the coaching team in the monitoring of players’ workloads during training sessions and matches.
4) Automatic detection and classification of teams’ formation – Team formation is a crucial part of any team sports’ strategy and it is one of many elements that combined, make a team successful. STATSports is using Unsupervised Machine Learning Classification methods (i.e. clustering algorithms) on spatial-temporal GNSS data to uniquely identify a team’s formation in and out of possession during a game. This allows the coaching team to understand how the team behaves during a match and decide if improvements can be made.