Publicação
Ticket price optimisation for a portuguese football club: The case of Sporting Clube de Portugal
| Resumo: | Ticket sales are one of the financial pillars of football clubs in Portugal. Over the last few years, interest in optimising ticket prices through dynamic pricing at sporting events has grown due to several successful applications in the United States. This work aims to create a data-driven pricing recommendation model tailored to the reality of Sporting Clube de Portugal. The multi-target model must be able to predict the demand for the 14 ticket categories in the club’s stadium, and, in a second phase, it must be able to find the optimal prices that maximise revenue based on the suggested prices. Data from 13 seasons, 286 games and five competitions were used to create a model that maximises the revenue obtained from ticket selling. A literature review was conducted on dynamic pricing in the sports industry, attendance forecasting and multi-target regression. Competition-specific and non-competition-specific models were tested, using ensembles of multi-target trees, ensembles of regressor chains (with linear regression and ensembles of single-target trees), and single-target approaches. To optimise the prices of a given match, the algorithm generates valid price combinations, which consider business-specific constraints, and uses a random search of the solution space to find the best price combination. Results show that there is no advantage in using approaches that consider the possible relationships between the various categories of tickets since the best model, considering an analysis of the mean absolute error, was the XGBoost with single-target trees, which is capable of predicting ticket sales for the matches from the Portuguese League. It was also inferred that a clear scope exists for setting better prices to increase revenue. To the authors’ knowledge, this is the first work that tests multi-target regression approaches to predict the number of tickets sold in any sport. It also addresses dynamic pricing in sports differently due to the constraints of Portuguese-specific regulations. Finally, it adds to the limited number of team-specific models and studies on dynamic pricing in European sports. |
|---|---|
| Autores principais: | Cymbron, António Maria Botelho de Sousa |
| Assunto: | Price Optimisation Sales Prediction Football Multi-Task Regression Machine Learning SDG 8 - Decent work and economic growth SDG 9 - Industry, innovation and infrastructure |
| Ano: | 2024 |
| País: | Portugal |
| Tipo de documento: | dissertação de mestrado |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | Universidade Nova de Lisboa |
| Idioma: | inglês |
| Origem: | Repositório Institucional da UNL |
| Resumo: | Ticket sales are one of the financial pillars of football clubs in Portugal. Over the last few years, interest in optimising ticket prices through dynamic pricing at sporting events has grown due to several successful applications in the United States. This work aims to create a data-driven pricing recommendation model tailored to the reality of Sporting Clube de Portugal. The multi-target model must be able to predict the demand for the 14 ticket categories in the club’s stadium, and, in a second phase, it must be able to find the optimal prices that maximise revenue based on the suggested prices. Data from 13 seasons, 286 games and five competitions were used to create a model that maximises the revenue obtained from ticket selling. A literature review was conducted on dynamic pricing in the sports industry, attendance forecasting and multi-target regression. Competition-specific and non-competition-specific models were tested, using ensembles of multi-target trees, ensembles of regressor chains (with linear regression and ensembles of single-target trees), and single-target approaches. To optimise the prices of a given match, the algorithm generates valid price combinations, which consider business-specific constraints, and uses a random search of the solution space to find the best price combination. Results show that there is no advantage in using approaches that consider the possible relationships between the various categories of tickets since the best model, considering an analysis of the mean absolute error, was the XGBoost with single-target trees, which is capable of predicting ticket sales for the matches from the Portuguese League. It was also inferred that a clear scope exists for setting better prices to increase revenue. To the authors’ knowledge, this is the first work that tests multi-target regression approaches to predict the number of tickets sold in any sport. It also addresses dynamic pricing in sports differently due to the constraints of Portuguese-specific regulations. Finally, it adds to the limited number of team-specific models and studies on dynamic pricing in European sports. |
|---|