Algorithm to predict top tennis player's shots before they play them

Source: Xinhua| 2019-01-22 13:53:06|Editor: Li Xia
Video PlayerClose

SYDNEY, Jan. 22 (Xinhua) -- Scientists have developed an intelligent algorithm which, using data from the Australian Open, is able to predict the next shot that the world's best tennis players will make.

Researchers from the Queensland University of Technology in Australia took data from the matches of Novak Djokovic, Rafael Nadal and Roger Federer in the early stages of the 2012 grand slam, and using their algorithm were able to predict the shots they would make in their finals matches.

Dr Simon Denman told Xinhua on Tuesday that to conduct the study the team used "Hawk-eye" data provided by Tennis Australia, the same system which helps officials determine whether a ball is in or out during play.

While not always hitting the mark, Denman said the algorithm does a pretty good job of latching on to a player's style.

"Our error is on average under a meter, so I think it's around 85-90 cm is the overall average error that we see," Denman said.

"There are some shots that are less accurate and there are some shots that are bang on."

Unlike a chair umpire, the algorithm tends to give players the benefit of the doubt - sometimes predicting a player will hit a shot in, when it actually goes wide or long.

"The model has got this understanding of effectively what the rules of the game are and where the ball should go and so it sort of corrects for those unforced errors that the players make," Denman explained.

Not all players were equally predictable - Federer proved the least consistent and perhaps the most versatile, with the predictions about his game returning the largest margin of error.

Unfortunately Denman said the algorithm cannot yet predict the overall winners for grand slam matches and tournaments.

"There is certainly the possibility to embed this research within a simulator to simulate matches and predict winners but as yet that hasn't been done," he said.

TOP STORIES
EDITOR’S CHOICE
MOST VIEWED
EXPLORE XINHUANET
010020070750000000000000011100001377650231