Player analytics for aspiring soccer players
December Labs partnered with Biostrap to conduct an AI-powered experiment to determine if their prosumer wearable devices could reliably classify and track soccer activities and provide athletes with valuable performance data.
Not too long ago, statistics in soccer meant mainly goals, assists, saves, corner kicks, penalty kicks and fouls. Times have changed thanks to companies specializing in sports analytics, player tracking technologies and the like, who have converted soccer metrics into a true data revolution. Player analytics is now a growing trend in amateur and professional sports. Athletes are intrigued by tracking their own performance, coaches at all levels embrace innovative techniques to improve their team’s overall performance on and off the field and major sports events such as the FIFA Soccer World Cup deliver metrics on players and teams in real-time direct to our screens - and we decided to get into the game as well, along with our client and partner Biostrap.
Biostrap is a next generation prosumer wearable platform focused on heart, sleep & activity monitoring. Its current offering includes activity specific data for Gym, Cycling, Cardio and Swimming workouts through the combined data measured by two complementary devices, a wristband and a shoe clip. Aware of the fact that soccer, the world's most popular sport, has been gaining traction in the United States in recent years, acclaimed to being the 3rd most favorite played team-sport and probably soon the 3rd favorite national sport to watch, Biostraps’ CEO & founder Sameer Sontakey was intrigued to pursue a concept of proof to determine if Biostrap devices could reliably classify and track soccer activities and provide athletes with valuable performance data.
In December of 2017, December Labs’ team of AI experts joined forces to conduct a sandbox environment test to classify soccer related movements by creating a neural network between the accelerometer and gyroscope of both Biostrap devices.
We conducted a field test, measuring acceleration and angle between consecutive samples of soccer related movements - strong and soft kicks with both legs, jump and run - to supply our soccer movement database with the necessary input, consisting of 35 motion-, orientation and rotation-related feature sets from the accelerometer and gyroscope readings for each movement.
Did it work?
Our initial findings concluded that:
- the wristband discriminated well between soft and strong kicks, but wouldn’t distinguish well between leg used and neither between jumping and running
- the shoe clip distinguished well which leg had been used and if someone jumped or ran, but had difficulties to differentiate between soft and strong kicks
Though each device successfully measured certain actions, neither devices gave us reliable enough results to classify all soccer related movements.
We then merged the features of both devices and validated results using a random forest classification to conclude if their complementary nature would result in a better outcome. Data reliability improved and indicated that with additional devices and a few tweaks we would be able to successfully classify the desired movements. So we started planning round two of testing, envisioning a new generation of devices providing aspiring athletes with invaluable performance-improving data.
In May of 2018 we went back into the field with the following alterations:
- we upgraded to using next-generation prototypes, enabling us to multiply the sampling frequency by 5, providing us with 300 samples per movement
- we added the variation of positioning both devices on the subjects leg, versus arm and leg as in the first round of testing
- we tested shorter ranges of movements
This time, we were successfully able to validate and classify strong and soft kicks with both legs, as well as differentiate between jumping and running when merging data from both complementary devices placed on the subjects legs for short ranges of movement (hurray!)
Biostrap is currently working on updating its hardware, conducting further testing with additional data and movement samples. The goal is to support soccer related movements by 2019 to enable professional (and aspiring) athletes to track their activity during a soccer match with their Biostrap devices, compare their performance to other players and keep track of personal improvements.
For us at December Labs, it was an inspiring experience to support our client on his path to product innovation and play a role in the continuous quest of providing athletes with relevant player analytics and insights.