Loading...

Application of Machine Learning Algorithms in Analysis and Prediction of Football Goalkeepers Performance and their Valuation in the Transfer Market

Zabihi, Matin | 2024

0 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 57661 (01)
  4. University: Sharif University of Technology
  5. Department: Industrial Engineering
  6. Advisor(s): Radman, Maryam
  7. Abstract:
  8. Football transformation into an industry with extraordinary financial flow has made this popular sport an attractive field for sports analysts. The development of the computer science and data science has led to the emergence of a concept called football match analysis. The goalkeepers, as a unique member of football teams and regrading their unique features and characteristics, play a crucial role in the outcome of competitions and the success of the teams. Therefore, analysis of various aspects of their performance is one of the essential issues in the application of data science in football. In this research, three separate parts have been studied. In the first part, by using pre-match factors and information, the performance indicators of the goalkeepers in the game, including Clean-sheet, the percentage of successful saves and the pass success rate (pass accuracy) are predicted using different machine learning models. In the second part, the performance measures and characteristics of the goalkeepers during each year were collected and their market value for the next transfer window is predicted by a regression model. In the third part of the study, by application of the two groups of goalkeepers’ performance indicators and based on the experience at the highest level of European club competition, we identified the differences in the performance measures and skills of these two groups of people by using machine learning algorithms.
    In the first part of the study, the importance and coefficients of variables were obtained in the machine learning models and then by statistical methods and test, their effect on the response variables was investigated. Two models, including Random Forest and Extreme Gradient Boosting, reported average accuracy of 79.4% and 79.9% for prediction of mentioned response variable, respectively. Also, in comparison with the baseline model (Decision Tree), Random Forest and Extreme gradient Boosting models showed improved accuracy in the first part. In the second part of the study, performance indicators related to ball distribution (passes) and defensive actions, were the most important measures, indicating the differences between the two groups of goalkeepers. Also, linear regression model for predicting market value of the goalkeepers while presenting 73% of R-Squared metric, showed that the focus of the pricing institutes is on team performance and the individual measures of the goalkeepers were ignored surprisingly in the market.
  9. Keywords:
  10. Performance Evaluation ; Machine Learning ; Talent Identification ; Football ; Football Players Valuation ; Football Goalkeepers

 Digital Object List

 Bookmark

No TOC