Bachelor Thesis BCLR-2020-92

BibliographyKurzendörfer, David: Tennis Match Outcome Prediction using LSTM Networks and Historical Averaging.
University of Stuttgart, Faculty of Computer Science, Electrical Engineering, and Information Technology, Bachelor Thesis No. 92 (2020).
51 pages, english.
Abstract

Abstract

In this thesis, a deep learning model has been created with the goal to predict tennis match outcome probabilities for ATP men's tennis matches. Leveraging state of the art architectures in sequence modelling and prediction power of hand engineered features, a model has been built combining Long-Short-Term-Memory (LSTM) and averaged statistics of historical tennis matches. New features have been created for example reflecting surprise in expectation of players when they tend to win as underdogs or lose as favorites. Likewise, an Elo rating model has been added to the set of features. The proposed model achieved 69.95% accuracy on a test set of 14770 tennis matches.

Department(s)University of Stuttgart, Institute of Parallel and Distributed Systems, Scientific Computing
Superviser(s)Pflüger, Prof. Dirk; Leiteritz, Raphael
Entry dateApril 9, 2021
   Publ. Computer Science