Publication: Análisis y predicción probabilística de resultados deportivos: tenis
Loading...
Identifiers
Publication date
2015-09-27
Defense date
2015-10-08
Authors
Advisors
Tutors
Journal Title
Journal ISSN
Volume Title
Publisher
Impact
Export
Abstract
Hoy en día, deporte, tecnología y apuestas son tres sectores con un gran crecimiento y que juegan un papel importante en la sociedad. En el presente proyecto uniremos los tres con el objetivo de desarrollar un modelo de predicción de resultados de partidos de tenis en los circuitos masculino y femenino que mejore el modelo previo existente.
Hablar de apuestas hoy en día puede parecer sencillo debido a la velocidad y rapidez con la que puedes acceder a ellas. Pero para plantearse un modelo de apuestas a largo plazo hay que contextualizar y analizar numerosos elementos que pueden afectarnos tales como el tipo de apuestas, las distintas casas que existen, las características del deporte al que vamos a apostar o los criterios que vamos a seguir, con el objetivo de obtener el máximo beneficio.
Para poder plantear un modelo con el cual podamos obtener beneficios a la hora de apostar, una vez decididos los factores mencionados anteriormente utilizaremos un enfoque Bayesiano y diferentes herramientas estadísticas a través del teorema de Bayes. Debemos seguir un determinado proceso para poder obtener parámetros que nos digan cómo se van a comportar los jugadores ante distintos partidos y así poder predecir el resultado de un encuentro de manera más sencilla. Este nuevo modelo que vamos a exponer tendrá como base técnicas como el aprendizaje máquina o la validación cruzada aplicadas sobre distintos factores determinantes en el tenis, con el fin de poder utilizar de la mejor manera toda la información a nuestra disposición que obtendremos de las distintas bases de datos. Con todo esto llegaremos al cálculo de la probabilidad que tienen los jugadores de ganar un encuentro, durante una serie de partidos de una o varias temporadas.
Para saber si los resultados de las probabilidades que obtenemos son buenos o no, utilizaremos funciones de coste. Debemos llegar al final del experimento obteniendo buenos resultados en esta parte, para luego poder apostar en una casa de apuestas online y comprobar el funcionamiento de nuestro modelo. El objetivo final es, como el de todo jugador, obtener algún tipo de beneficio.
Mucho se escribe sobre apuestas, estrategias, criterios y distintas formas de apostar. Nosotros estamos convencidos de que con el modelo riguroso que vamos a presentar los resultados, tanto a nivel cualitativo como cuantitativo, serán muy interesantes.
Nowadays, sports, technology and betting are three sectors with high growth and which play an important role in our society. In this project we will join the three together in order to develop a model for predicting outcome of tennis matches in the male and female tours to improve the existing previous model. Speaking of betting today may seem easy due to the speed and quickness with which you can access them. However, to propose a long-term betting model, we must contextualize and analyze many elements that can affect such as the type of betting, the different sports bookmakers that exist, the characteristics of the sport that we will bet or the criteria that we will follow, in order to obtain the maximum benefit. To propose a model with which we can obtain benefits when betting, once determined the factors mentioned above we will use a Bayesian approach and different statistical tools through the Bayes theorem. We must follow a certain process to obtain parameters that tell us how they will behave players to different matches and thus to predict the outcome of a match more easily. This new model that we will expose is technically based on the machine or cross-validation learning applied to different determinants in the tennis, in order to be able to use in the best way all the information at our disposal, what we will gain from the different bases of data. With all this, we will come to the calculation of the probability that players win a game, during a series of matches one or more seasons. To find out if the results of the chances we get are good or not, we will use cost functions. We must get to the end of the experiment performing well in this part, then to to bet in online sports bookmarkers and finally to check the functioning of our model. The ultimate goal is, like every player, get some benefit. Much is written about betting, strategies, criteria and other forms of betting. We are convinced that with the rigorous model that we will present the results, both qualitatively and quantitatively, will be very interesting.
Nowadays, sports, technology and betting are three sectors with high growth and which play an important role in our society. In this project we will join the three together in order to develop a model for predicting outcome of tennis matches in the male and female tours to improve the existing previous model. Speaking of betting today may seem easy due to the speed and quickness with which you can access them. However, to propose a long-term betting model, we must contextualize and analyze many elements that can affect such as the type of betting, the different sports bookmakers that exist, the characteristics of the sport that we will bet or the criteria that we will follow, in order to obtain the maximum benefit. To propose a model with which we can obtain benefits when betting, once determined the factors mentioned above we will use a Bayesian approach and different statistical tools through the Bayes theorem. We must follow a certain process to obtain parameters that tell us how they will behave players to different matches and thus to predict the outcome of a match more easily. This new model that we will expose is technically based on the machine or cross-validation learning applied to different determinants in the tennis, in order to be able to use in the best way all the information at our disposal, what we will gain from the different bases of data. With all this, we will come to the calculation of the probability that players win a game, during a series of matches one or more seasons. To find out if the results of the chances we get are good or not, we will use cost functions. We must get to the end of the experiment performing well in this part, then to to bet in online sports bookmarkers and finally to check the functioning of our model. The ultimate goal is, like every player, get some benefit. Much is written about betting, strategies, criteria and other forms of betting. We are convinced that with the rigorous model that we will present the results, both qualitatively and quantitatively, will be very interesting.
Description
Keywords
Modelos probabilísticos, Apuestas, Validación cruzada, Probabilidad, Juegos de azar, Tenis