Rouse Award: Sophie Chiang’s research into table tennis match prediction
Sophie Chiang (Upper Sixth) investigated the extent to which machine learning could be used to predict the outcome of table tennis matches for her Rouse Award-winning project.
A former national junior table tennis champion, Sophie decided to combine her passion for the sport with her interest in computer science and maths for her research.
She said: “It resulted in me looking into a problem which hasn’t really been tackled before, so my project was focused on bridging the gap between data collection and predictive analysis in table tennis.
“For prediction, you need a lot of data and information. Due to recent developments in technology, such as multi-class event spotting and object tracking, it has meant this data can now be obtainable from these matches automatically.
“This is more reliable than manually counting, for example, how many times the ball goes over the net, because table tennis is very fast-paced and if it’s done by computer, it’s more reliable.”
Sophie began by reading more deeply into machine learning – a branch of artificial intelligence in which models are built that learn patterns and relationships from inputted data.
She then took statistics collated from the Tokyo 2020 Olympics table tennis tournament and the professional German Table Tennis League, such as the percentage of points won on or against serve by individual players in matches, to work on her project.
“I had to collect the data in a readable, structured format and then feed it into various machine learning models to predict who would win or lose,” said Sophie.
“Within each match, there’s a lot of statistics and information you can use and I extracted the information I thought was useful.”
Sophie also developed her own features – one taking account of player rankings, another considering players’ all-round ability from backhand and forehand strength to their capability in long or short rallies – for consideration.
She found the best machine learning model provided around 70% accuracy in predicting the outcome of the matches from the above competitions.
Sophie, who plans to study computer science at university, said: “I think this kind of thing is used in other sports by coaches to analyse players’ performances and see where they have to improve in a particular match so it could be quite useful.
“I loved doing this project. I was quite surprised to win the Rouse Award, but I really enjoyed the interview with the judging panel and it was nice to talk about something I’d found so interesting.”
Published