Study on the decision of sports lottery betting-Evidence from National Basketball Association games

• Main Authors: CHING,PO-YEN, 慶柏延
• Other Authors: SUN, MEI-JUI
• Format: Others
• Language: zh-TW
• Published: 2016
• Online Access: http://ndltd.ncl.edu.tw/handle/m4pnha

碩士 === 東吳大學 === 國際經營與貿易學系 === 104 === It is obvious to see that betting market for Taiwan sports lottery is growing year by year, from the sales of 5.2 billion NTD in 2008 to 24 billion NTD in 2014, Due to the general public or sport lottery fans judging the game intuitively, then they have the high probability to lose. This study thinks that sports lottery would be a technical, highly predictable target of investment, and suggests, it could produce a great return via the investment management, data analysis, risk assessment and appropriate strategy application for sport lottery. The goal of this study is providing models and strategies of statistical analysis through gathering of game records and relevant information, in order to increase winning bets for general public or sport lottery fans. This study uses the NBA games records from 2012 to 2015 to analyze the game results which are the winners (the home team or the visitor) This study collected the data from 2012 to 2014, a total of 2459 games to create a forecast model, and applied 1230 games scores from 2014 to 2015 as a testing sample to predict a total of 719 games from 2015 to 2016, in order to provide a effective betting strategy by the comparison of virtual betting and intuitive betting results. The research not only uses the data of game scores but also adding the factors such as team advantages and team combination in the model. It applies the logistic regression analysis to filter out important variables that would affect the game results, then through neural network to create four different sets of forecast models, which forecast high winning percentage (low odds), low winning percentage (high odds), and home matches etc. As a result of the prediction, variable filter has better predictive capability than non variable filter. From the perspective of the betting option, selecting low winning bets predict the accuracy of 83.5%, where the return on investment could reach to 115.33%.