Australian Institute of Sport taps into data to help athletes train for gold

The Australian Institute of Sport has been using data analytics to ensure some of Australia's top athletes have been able to train at their optimum level for the Rio Olympics and Paralympics.
Written by Aimee Chanthadavong, Contributor

The Australian Institute of Sport (AIS) has predicted, and is hopeful, the Australian team will pick up six or seven gold medals at the upcoming Rio Olympics.

The AIS is able to make this prediction because the organisation has been working daily with professional athletes, who have been training for the Rio Olympics and Paralympics.

A key part of working with athletes has been a three-year-long pilot of collating data and analysing it to reduce the likelihood of an athlete becoming injured or falling ill. The information is collected from 2,000 athletes each week, including 300 data points per athlete, making up a total of 600,000 data points per week.

Specifically, the collected data includes looking at how much the athlete has trained, how they felt, how well they slept, their psychological data, and physiotherapy information.

According to Nick Brown, AIS performance science and innovation deputy director, using data and analytics means athletes are able to train and compete consistently, without losing days to recovery or illness.

"We normally lose up to 20 percent of training time because an athlete is sick or injured, and therefore are unable to meet their performance goal," he said.

AIS has partnered with Microsoft and BizData to use predictive analytics and machine learning to analyse the collected data, which is uploaded each night through an Azure SQL Database to the Athlete Management System, where all of the data is stored. The custom algorithms are then updated at 6am every morning to enable the coaches to modify the training for the day according to the data.

While the project is still in pilot phase, Joseph Winter, AIS innovation, research and development head of discipline, said some of the early wins to date include being able to predict which athletes will be injured in the next three days.

The pilot has also forced the company to change the way it manages and handles data. Winter said that previously, athletes' data was disparate and placed across multiple hard drives, but having the management system means all of the data is kept in a centralised database.

At the same time, he said the organisation has started to learn how to manage data in an ethical way, given it is made up of athletes' personal information, and therefore requires the AIS to be more compliant around privacy policies.

"We have recently created policies around how you access the data ... what the thresholds are, because if a research paper is published, you don't want someone to be able to identify the data so you have to have a lot of discussions," he said, noting all the data is currently kept in-country.

Winter added there are plans to create a similar database for the science arm of the organisation, which he believes will be another big task for the AIS.

"We have the athlete management system in quite a good model. We need to somehow bring the science house along and get them to realise that the data is not just their data for their [research] paper; it's an asset of the Institute of Sport. Somehow we have to start to create a common science database for our research," he said.

Going forward, Brown believes that as the AIS becomes flooded with more data, it will be a task of trying to decipher between what data is most relevant.

"There's going to be a balance of discovery and new data because it's just increasing all the time with wearable sensors, Internet of Things; we're getting more and more data off athletes and coaches. We need to balance what's new and useful ... versus what's something we can tangibly use," he said.

Editorial standards