Decision Support in Racket Games

Thumbnail Image
Date
2022-12
Authors
Kumari, Anju
Major Professor
Mitra, Simanta
Prabuh, Gurpur
Advisor
Committee Member
Chou, Li-Shan
Journal Title
Journal ISSN
Volume Title
Publisher
Authors
Research Projects
Organizational Units
Journal Issue
Is Version Of
Versions
Series
Department
Abstract
In order to prepare for the next match, players examine enormous amounts of footage throughout practice to determine a potential opponent's patterned behavior and strategic tendencies. However, viewing and analyzing an opponent’s previous matches takes a lot of time, so the players often end up analyzing only a few or recent matches of the potential opponent. This clearly isn't optimal in terms of predictive analytics because the match they're looking at may not be a representative match of their opponent. Many important factors that contribute to players' playing style might get overlooked by tennis players in the given timeframe. This may lead to misinterpretation of the opponent's important strategic elements or the opponent's likely playing style in subsequent matches. The racket games lack the use of technology to assist tennis players in learning the game and help them improve despite its popularity. And, there is an enormous amount of data generated in every professional racket game in the form of vision-based tracking data and also text-based data. Given the abundance of historical tennis data, players' characteristics and match characteristics might all be combined to create a set of labeled training examples. This project explores the application of supervised machine learning algorithms to provide the decision support for the tennis players by recommending them the shot type and serve direction that the opponent is most likely to hit in a particular context of the game. The machine learning models like Random Forest, KNN, SVC and also the latest classification algorithms like Catboost are used to categorize the features and predict the outcome.
Comments
Description
Keywords
Citation
DOI
Source
Copyright
2022