Publication: Sistema de predicción de resultados en eventos deportivos y su aplicación en las apuestas
Loading...
Identifiers
Publication date
2013-06
Defense date
2013-07-02
Authors
Advisors
Tutors
Journal Title
Journal ISSN
Volume Title
Publisher
Impact
Export
Abstract
El Proyecto de Final de Carrera que se presenta puede dividirse en dos partes: la primera
de ellas tiene como principal objetivo la creación de un Sistema de Predicción capaz de
predecir el resultado más probable de un partido en un determinado evento deportivo.
Después de desarrollar el Sistema de Predicción, se pasará a la segunda parte del
proyecto, que será la encargada de la explotación de estas predicciones para obtener
beneficios en las casas de apuestas. En la fase de explotación, se han utilizado tanto
estrategias que son ya utilizadas por los aficionados a las apuestas, como nuevas
estrategias que se han ideado para explotar al máximo las características del Sistema de
Predicción que se ha desarrollado en la primera fase.
El sistema se centra en la predicción de resultados de dos deportes: fútbol (englobando
varias competiciones tanto nacionales como internacionales) y baloncesto (considerando
únicamente la liga americana de baloncesto, también conocida como NBA).
Para el desarrollo del Sistema de Predicción, que estará basado en clasificadores, se han
recogido datos de un gran número de partidos de los deportes y competiciones nombrados
anteriormente. A través de un proceso de minería de datos se ha determinado cuáles son
los atributos que aportan mayor información para realizar la predicción.
Una vez se tienen todos los datos correctamente etiquetados, se ha utilizado la
herramienta WEKA para crear diversos modelos de predicción basados en la
clasificación. A través de esos modelos generados y tras recoger otra tanda de datos de
los eventos deportivos, se ha analizado que modelos eran los que obtenían mejores tasas
de acierto.
Para automatizar la tarea de predicción de los resultados, se ha creado un archivo Excel,
con el que a través de varias macros que implementan los clasificadores seleccionados se
calcula el resultado más probable del partido y el riesgo estimado que ese resultado lleva
asociado.
Posteriormente, se definirán las estrategias que se van a seguir para intentar obtener
beneficios en las casas de apuestas a través de las predicciones realizadas por el Sistema
de Predicción. Entre estas estrategias se ha desarrollado un Algoritmo Genético, que a
partir de los datos ofrecidos por el Sistema de Predicción es capaz de conseguir una
combinación de apuestas para las que se maximiza el beneficio minimizando el riesgo.
Tras definir las estrategias e implementar el algoritmo en el que se basará una de ellas, se
realizará un estudio de beneficios en el que se podrá ver qué estrategias son capaces de
generar beneficios. Además, se podrá comprobar qué competiciones son las que mejores
tasas de aciertos tienen y las que más beneficios son capaces de generar.
This project can be divided into two parts: the first one has as main objective the creation of a Prediction System capable of predicting the most likely result of a match in a particular sport event. After developing the Prediction System, we will go to the second part of the project, which will be responsible for the exploitation of these predictions for profit in betting. In the operational phase, will be used both strategies that are already used by the betting fans, as new strategies that have been designed to exploit the characteristics of the Prediction System which has been developed in the first phase. The system focuses on two sports: football (considering several national and international competitions) and basketball (considering only American Basketball League, also known as NBA). To develop the Prediction System, which will be based on classifiers, we have collected data from a large number of sports matches and competitions listed above. Through a data mining process, it has been determined what attributes provide more information for the prediction. Once we have all the data correctly labeled, we used the WEKA tool to create different prediction models based on classification. Through these models generated and after picking up another round of sporting event data, it has been analyzed what models has the best hit rates. To automate the task of prediction, we have created an Excel file, which through several macros which implement the selected classifiers, calculate the most likely result of the match and the estimated risk that is associated with that result. Later, we will define the strategies that are going to follow to try to make profit in betting through the predictions made by the Prediction System. For one of these strategies we have developed a Genetic Algorithm, which is based on the data provided by the Prediction System, and is able to get a combination of bets for maximizing profit while minimizing risk. After defining the strategies and implement the algorithm in which one of them will be based, we will make a study of benefits in which we can see which strategies are able to generate profits. Additionally, we can see which competitions have the best hit rates and the best profit ratio.
This project can be divided into two parts: the first one has as main objective the creation of a Prediction System capable of predicting the most likely result of a match in a particular sport event. After developing the Prediction System, we will go to the second part of the project, which will be responsible for the exploitation of these predictions for profit in betting. In the operational phase, will be used both strategies that are already used by the betting fans, as new strategies that have been designed to exploit the characteristics of the Prediction System which has been developed in the first phase. The system focuses on two sports: football (considering several national and international competitions) and basketball (considering only American Basketball League, also known as NBA). To develop the Prediction System, which will be based on classifiers, we have collected data from a large number of sports matches and competitions listed above. Through a data mining process, it has been determined what attributes provide more information for the prediction. Once we have all the data correctly labeled, we used the WEKA tool to create different prediction models based on classification. Through these models generated and after picking up another round of sporting event data, it has been analyzed what models has the best hit rates. To automate the task of prediction, we have created an Excel file, which through several macros which implement the selected classifiers, calculate the most likely result of the match and the estimated risk that is associated with that result. Later, we will define the strategies that are going to follow to try to make profit in betting through the predictions made by the Prediction System. For one of these strategies we have developed a Genetic Algorithm, which is based on the data provided by the Prediction System, and is able to get a combination of bets for maximizing profit while minimizing risk. After defining the strategies and implement the algorithm in which one of them will be based, we will make a study of benefits in which we can see which strategies are able to generate profits. Additionally, we can see which competitions have the best hit rates and the best profit ratio.
Description
Keywords
Diseño de software, Apuestas deportivas, Minería de datos, Algoritmos genéticos, Predicción