Fecha
2025
Editor/a
Director/a
Tutor/a
Coordinador/a
Prologuista
Revisor/a
Ilustrador/a
Derechos de acceso
info:eu-repo/semantics/openAccess
Título de la revista
ISSN de la revista
Título del volumen
Editor
Universidad Nacional de Educación a Distancia (UNED). E.T.S. de Ingeniería Informática (UNED), 2025
Citas
0 citas en 
0 citas en 
Resumen
Este trabajo tiene como objetivo analizar la predicción de los resultados de los partidos de voleibol a nivel profesional mediante el uso de métricas de habilidades técnicas para aclarar si es necesaria la inclusión de métricas avanzadas, tal como ocurre en otros deportes de élite. Para ello, mediante el uso de técnicas multivariantes, se ha construido un conjunto de modelos de predicción con datos de la PlusLiga polaca entre los años 2008 y 2023. Se proponen dos enfoques: enfoque individual, en el cual solo se tienen en cuenta las estadísticas de un solo equipo, sin importar lo que haga el equipo rival, y enfoque comparativo, en el cual se tienen en cuenta las estadísticas de ambos adversarios, proponiéndose como variables las diferencias de dichas estadísticas. En cada enfoque se crean dos modelos: un modelo de regresión logística en el que solo se utilizan habilidades técnicas y otro modelo bayesiano jerárquico, que tiene en cuenta efectos aleatorios por equipos más allá de las habilidades técnicas. Los resultados muestran un buen rendimiento de los modelos para predecir la victoria. Al hacer una evaluación específica de la duración de los partidos (tres, cuatro o cinco sets), se evidencia el excelente rendimiento que ofrecen para partidos de tres sets. En los partidos de cuatro sets, el rendimiento de los modelos desciende levemente, y en los que se juegan a cinco sets la capacidad predictiva aún decae más. Para examinar la validación de los modelos en otras ligas de élite, se utilizaron datos de partidos de la Superliga rusa masculina de la temporada 2024-2025, ofreciendo resultados similares o superiores a los obtenidos con los datos de la liga polaca. Del enfoque individual se pudo conseguir un ranking de importancia de métricas técnicas, destacando en este orden los ataques con remates, las defensas o digs, y los bloqueos con puntos. En el enfoque comparativo se consiguieron reducir las variables predictivas a solo dos: diferencias de la eficiencia en ataque y diferencias de la eficiencia en saque; pudiéndose de esta manera visualizar la naturaleza lineal del modelo. También se analizó la influencia sobre la victoria que tiene una nueva variable, Blk_As (bloqueo que posibilita una jugada de contraataque), llegando a la conclusión de que no hay evidencias de que sea una variable relevante en lo que a victoria se refiere. Se propusieron mejoras como la obtención de conjuntos de datos en que las estadísticas de los partidos están separadas por sets para construir modelos con mejor rendimiento. Finalmente, se llegó a la conclusión de que, a pesar del buen desempeño de los modelos, sería muy positiva la inclusión de métricas avanzadas, no solo para predecir el resultado final de un partido, sino para tener más conocimiento y entender lo que ocurre en el juego, que no puede ser explicado con métricas básicas.
This work aims to analyze the prediction of professional volleyball match outcomes using technical skill metrics to assess whether the inclusion of advanced metrics—common in other elite sports—is necessary. To achieve this, a series of predictive models were built using multivariate techniques and data from the Polish PlusLiga between 2008 and 2023. Two approaches are proposed: an individual approach, which only considers the statistics of one team regardless of the opponent, and a comparative approach, which incorporates the statistics of both teams by using the differences between them as predictive variables. For each approach, two models were created: a logistic regression model that uses only technical skills, and a hierarchical Bayesian model that includes random effects by team beyond those technical metrics. The results show that the models perform well in predicting match victories. When evaluating performance based on match length (three, four, or five sets), the models showed excellent results for three-set matches. In four-set matches, model performance slightly decreased, and in five-set matches, predictive accuracy declined even further. To test the generalizability of the models to other elite leagues, match data from the Russian men’s Superliga during the 2024–2025 season was used, yielding results that were similar or even superior to those obtained with Polish league data. From the individual approach, it was possible to generate a ranking of the most important technical metrics: attack kills, digs, and points from blocks stood out in that order. In the comparative approach, the predictive variables were reduced to just two: attack efficiency and serve efficiency, clearly revealing the model’s linear nature. The influence of a new variable, Blk_As (block enabling a coun- terattack), was also analyzed, and it was concluded that there is no evidence suggesting it is relevant to match outcomes. Improvements were proposed, such as obtaining datasets where match statistics are broken down by set, in order to develop higher-performing models. Finally, we concluded that, despite the strong performance of the models, the inclusion of advanced metrics would be highly beneficial—not only to better predict match outcomes but also to enhance understanding of what happens during a game, beyond what basic metrics can explain.
This work aims to analyze the prediction of professional volleyball match outcomes using technical skill metrics to assess whether the inclusion of advanced metrics—common in other elite sports—is necessary. To achieve this, a series of predictive models were built using multivariate techniques and data from the Polish PlusLiga between 2008 and 2023. Two approaches are proposed: an individual approach, which only considers the statistics of one team regardless of the opponent, and a comparative approach, which incorporates the statistics of both teams by using the differences between them as predictive variables. For each approach, two models were created: a logistic regression model that uses only technical skills, and a hierarchical Bayesian model that includes random effects by team beyond those technical metrics. The results show that the models perform well in predicting match victories. When evaluating performance based on match length (three, four, or five sets), the models showed excellent results for three-set matches. In four-set matches, model performance slightly decreased, and in five-set matches, predictive accuracy declined even further. To test the generalizability of the models to other elite leagues, match data from the Russian men’s Superliga during the 2024–2025 season was used, yielding results that were similar or even superior to those obtained with Polish league data. From the individual approach, it was possible to generate a ranking of the most important technical metrics: attack kills, digs, and points from blocks stood out in that order. In the comparative approach, the predictive variables were reduced to just two: attack efficiency and serve efficiency, clearly revealing the model’s linear nature. The influence of a new variable, Blk_As (block enabling a coun- terattack), was also analyzed, and it was concluded that there is no evidence suggesting it is relevant to match outcomes. Improvements were proposed, such as obtaining datasets where match statistics are broken down by set, in order to develop higher-performing models. Finally, we concluded that, despite the strong performance of the models, the inclusion of advanced metrics would be highly beneficial—not only to better predict match outcomes but also to enhance understanding of what happens during a game, beyond what basic metrics can explain.
Descripción
Categorías UNESCO
Palabras clave
curva ROC, habilidades técnicas, métricas avanzadas, modelo bayesiano jerárquico, regresión logística, voleibol, advanced metrics, hierarchical bayesian model, logistic regression, ROC curve, technical skills, volleyball
Citación
Soler Rocha, Sergio. Trabajo Fin de Máster: "Análisis de la relación entre métricas de habilidades técnicas y resultado en voleibol: un estudio con técnicas multivariantes". Universidad Nacional de Educación a Distancia (UNED), 2025
Centro
E.T.S. de Ingeniería Informática
Departamento
Inteligencia Artificial