24th EANN 2023, 14 - 17 June 2023, León, Spain

Subsampled dataset challenges and Machine Learning techniques in Table Tennis

Dimitrios Simopoulos, Andreas Nikolakakis, George Anastassopoulos


  Artificial Intelligence has been widely used almost in every aspect of our lives, and sports activity is no exception. In that field, table tennis is a very demanding sport, in which predicting the winner of a point or a match can be quite challenging. This paper explores the application of several machine learning algorithms in subsampled table tennis data. Especially, the Multilayer Perceptrons, Random Forests, and Gradient Boosted Trees have been used in order to predict the result of individual points and matches. The algorithms were trained on real data from official First Division matches of the Hellenic Table Tennis Federation. The Gradient Boosted Trees achieved the highest level of accuracy (98.36%) in the prediction of each individual point and (74.92%) in the prediction of the winner of the match. The 98.36% accuracy is very high, while the 74.92% is affordable, because the match winner could have won even fewer points than its opponent.Also, Gradient Boosted Trees achieved the less computational time, making it suitable for use in real-time systems to assist tactical coaching in table tennis.  

*** Title, author list and abstract as seen in the Camera-Ready version of the paper that was provided to Conference Committee. Small changes that may have occurred during processing by Springer may not appear in this window.