AI beats goalkeepers at predicting which way penalty taker will shoot
Goalkeepers find it difficult to guess in which direction a penalty lessee will draw
Javier Soriano / AFP via Getty Images
The in -depth learning models trained on more than 1000 penalty kicks in football matches can predict in which direction the ball will be better than real goalkeepers.
“The penalty kicks are among the most decisive moments in football, often determining the outcome of major tournaments,” explains David Freire-Obregón at the University of Las Palmas of Gran Canaria, Spain. “Despite this, real -time support for goalkeepers is still largely based on intuition. We wanted to explore if automatic learning could predict the management of the body movement of a botter.”
Freire-Obregón and his colleagues therefore scratched 1010 kicks of real television matches in Spain. Among these clips, 640 were deemed analyzed by AI models, while the others were thrown to be blurred, too short or obstructed.
Each clip was then fed with 22 deep learning models, which should guess if the penalty would go to the left, right or in the middle, depending on the video sequences and the simple fact of whether the player was right or left.
The most efficient model was able to identify properly if the ball went to the right, on the left or in the middle of 52% of the time – better than the precision of 46% of real goalkeepers in the matches. When the researchers deleted the average option less used, the precision of the model increased to 64%, or nearly 10 percentage points higher as human goalkeepers given the same information.
The researchers were surprised “by the quantity of indices of subtle movement, even before the ball is kicked, can reveal the intention,” says Freire-Obregón. He hopes that the information can be useful for goalkeepers for training, but the use of AI prediction in a match situation will be more a challenge.
“What we aim to explore afterwards is if these probabilities can be anticipated in advance for the penalty kicks, using only the botter movement before the blow,” he said. “And in the affirmative, how much forecast can be made while maintaining significant precision.”
Subjects:
