| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 6.94 MB | Adobe PDF |
Authors
Advisor(s)
Abstract(s)
O mercado do futebol está em alta, com jogadores e treinadores sendo cada vez mais valorizados. Para garantir um desempenho superior, é crucial fazer escolhas criteriosas na contratação. Além disso, há uma demanda crescente por dados nesse setor, e métricas avançadas, como "expected goals", estão a tornar-se populares na análise de jogos de futebol. Essas métricas, originalmente usadas por mercados de apostas, agora são adotadas por comentadores e treinadores renomados. Isso indica que a análise de dados é essencial para melhorar o desempenho de todos os envolvidos no futebol. Diante desse cenário, surge a necessidade de desenvolver uma solução que consiga explorar sequências e padrões de jogo através de análises avançadas e consiga extrair padrões de jogo a partir de imagens de sequências. A metodologia utilizada neste projeto de pesquisa é a Design Science Research. Inicialmente, foi realizada uma revisão bibliográfica sobre os tipos de dados existentes no contexto do futebol, as métricas avançadas atualmente em alta no mundo analítico desportivo e soluções existentes no ramo. Foram identificadas e descritas algumas das características e limitações mais comuns dos serviços atuais do mercado. Este trabalho pretende apresentar uma proposta que inove no cálculo da métrica de xG, consiga identificar diversas estatísticas calculadas a partir de dados de eventos e consiga estabelecer uma relação entre esses dados, as sequências das equipas e o estilo de jogo da equipa. O sistema Verance App utiliza dados do tipo de fluxo de eventos para calcular estatísticas para todas as equipas que atuaram nas principais 6 ligas durante a presente temporada (2022/23) e apresentar estatísticas de todas as sequências e ações destas mesmas equipas. Para além disto, apresenta também a funcionalidade de apresentação das 3 equipas mais semelhantes em análise. A Verance App não foi utilizada por nenhuma equipa real para fornecer informação de melhoria dos resultados desportivos, mas foi avaliada tendo em conta os seus 3 componentes principais, o modelo xG, o modelo xT e a componente de extração dos padrões das sequências. A análise confirma que a solução projetada, na maioria das circunstâncias, apresenta resultados superiores aos dos serviços atuais do mercado.
The football market is booming, with players and coaches being increasingly valued. To ensure superior performance, it is crucial to make careful choices in recruitment. Additionally, there is a growing demand for data in this industry, and advanced metrics like "expected goals" are becoming popular in football analysis. These metrics, originally used by betting markets, are now adopted by renowned commentators and coaches. This indicates that data analysis is essential for improving the performance of everyone involved in football. In this context, there is a need to develop a solution that can explore game sequences and patterns through advanced analysis and extract patterns from sequence images. The methodology used in this research project is Design Science Research. Initially, a literature review was conducted on the types of data available in the context of football, the currently popular advanced metrics in sports analytics, and existing solutions in the field. Some of the common characteristics and limitations of current market services were identified and described. This work aims to propose an innovative approach to calculating the xG metric, to identify various statistics derived from event data, and to establish a relationship between this data, team sequences, and team playing style. The Verance App system utilizes event stream data to calculate statistics for all teams participating in the top 6 leagues during the current season (2022/23) and presents statistics for all sequences and actions of these teams. Additionally, it provides the functionality to present the top 3 most similar teams for analysis. The Verance App has not been used by any real team to provide performance improvement insights. However, it has been evaluated based on its three main components: the xG model, the xT model, and the sequence pattern extraction component. The analysis confirms that, in most circumstances, the designed solution outperforms the current market services.
The football market is booming, with players and coaches being increasingly valued. To ensure superior performance, it is crucial to make careful choices in recruitment. Additionally, there is a growing demand for data in this industry, and advanced metrics like "expected goals" are becoming popular in football analysis. These metrics, originally used by betting markets, are now adopted by renowned commentators and coaches. This indicates that data analysis is essential for improving the performance of everyone involved in football. In this context, there is a need to develop a solution that can explore game sequences and patterns through advanced analysis and extract patterns from sequence images. The methodology used in this research project is Design Science Research. Initially, a literature review was conducted on the types of data available in the context of football, the currently popular advanced metrics in sports analytics, and existing solutions in the field. Some of the common characteristics and limitations of current market services were identified and described. This work aims to propose an innovative approach to calculating the xG metric, to identify various statistics derived from event data, and to establish a relationship between this data, team sequences, and team playing style. The Verance App system utilizes event stream data to calculate statistics for all teams participating in the top 6 leagues during the current season (2022/23) and presents statistics for all sequences and actions of these teams. Additionally, it provides the functionality to present the top 3 most similar teams for analysis. The Verance App has not been used by any real team to provide performance improvement insights. However, it has been evaluated based on its three main components: the xG model, the xT model, and the sequence pattern extraction component. The analysis confirms that, in most circumstances, the designed solution outperforms the current market services.
Description
Keywords
xG xT Futebol Estatísticas Avançadas Inteligência Artificial Aprendizagem Profunda Python Pytorch Mediator React Soccer Advanced Stats Artificial Intelligence Deep Learning