Basketball research automates play patterns to compare team performance

New analysis of elite women’s basketball automatically determines a team’s chances of scoring high or low even though the trajectory of the ball looks similar, in research developed by QUT data scientists.

The study, recently published in PLEASE ONEusing existing data from 72 international women basketball games ranging from the 2014 FIBA ​​World Championships to the 2016 Rio Olympics.

The results provide insights to assist coaches in scrutinizing teams’ effective or problematic play by categorizing and tracking the dynamics of ball movements to see if a team whether or not to score and the types of play related to the scoring results.

Although data on spatial characteristics such as basketball bounce, speed, ball possession time, points scored as well as previous team histories are varied, the study attempts to group the The style of play using the movement of the ball is limited.

This new study, applying the concept of ‘dynamic time curvature’, was led by Professor Kerrie Mengersen and Dr Paul Wu from QUT’s Data Science Centre, Dr Wade Hobbs from the Australian Institute of Sport, University of Sydney and QUT. Student Alan Yu.

Dr Wu said the study was motivated by questions about the unpredictability of plays and whether that leads to better scoring outcomes.

He said that by building on that data and incorporating the trajectory of the ball, the researchers were able to investigate how the patterns appeared based on the data.

“If we were to take an identical round of basketball, do it at the same speed, but start the second round a second later than the second, this would be like a completely different turn to the machine. calculation, but people see the same. play,” he said.

“Dynamic Time Warping provides a way to map one trajectory to another to better assess how similar they are automatically.

“This is a way to organize many, many hours of video recording intended to help coaches and athletes identify key strengths and weaknesses to review and highlight something unclear, such as if a team favors one side of the court.”

The data also unsurprisingly found faster breaks and larger ball movements, especially changes in the direction of the ball. ball movement contributes to a higher scoring rate.

The project opens up opportunities for QUT data science student Alan Yu participated in a “study vacation”.

“I found the experience challenging but also rewarding, especially when it comes to developing easy-to-understand visualizations to tell stories to people interested in basketball or data science,” he said.

“Without much prior knowledge of basketball, I found the results very interesting and surprising.

“Often you’ll hear coaches and experts analyze the games and the different tactics and preferences the teams may have, but this puts a quantifiable touch to understanding the sport. This is better from a systematic point of view.

“It’s amazing to see similar plays with very different results when performed by different teams at different speeds.”

Professor Mengersen said the study is a prime example of what is achieved through partnerships between researchers and industry practitioners.

“New methods have been developed to address an important real-world question, and new insights will feed back directly to improve the sport industry,” she said.

“This two-way study benefits everyone and has implications in both sport and new knowledge.”