| Sports Betting: MLB | ||||||||||||||||||||||||||||||||||||||||
|
More than any of the sports I have studied, I probably have the best sense of this game. It is also the most data intensive, so seemingly creates the best opportunity to really breakdown the sport. However, this data intensity also creates the need for robust datasets, the kind beyond my abilities in Excel. I did assemble a fairly thorough dataset (data fields summarized below) but it was only for 220 games. I hope some day to assemble a season's worth of data and see if I can find the relationships I need to outperform the line. I did get a friend to do some database programming to get the data I need, but he took a job in the middle of the project, so it stalled. Winning Drivers Based on my analysis of baseball, I know OPS is a very helpful metric to determine the quality of baseball team. My basic logic was to compare each team's OPS of the hitting and the OPS allowed of the pitching and see if the net result had any predictive value in the outcomes. For the pitching, I inserted the data for the specific pitcher and tallied the cumulative data for the relief corp of each team. If I could find some useful data, I could then compare the predictive values against the implied expected winning percentage in the line. I expanded these variables to include winning percentage and the opinion of people who track the games. Assembling this data set was very labor intensive. It involved, each day, cut & pasting from the web the team OPS (team hitting, individual pitching data). The table below summarizes all the variables I created, assembled into a data set and ran against the "net OPS" outcome of the game. My logic was to assemble an intuitive set of variables and then run the data set to see which ones provided the best model. Pitching Variables
These variables above were further enhanced to account for home vs. road splits as well as the effects of the ballpark where the game was being played. I would do the same calculations for both home and visitor, and derive a "net OPS pitching" of the two teams, for basic, split adjusted and split adjusted that factors in the ballpark. Hitting Variables
Winning Percentage/Winning Metrics
Third Party Variables A new, very helpful site launched recently (www.wagerline.com), that is a very helpful source for game information, especially odds. In addition, they track the picks of their users. I decided to track this data to see if it would help in the model as well.
Results For all this work, I did not get much. Counter-intuitive relationships, little explanatory value. My data set was only 220 games. One of ideas was to build this data set out, for an entire season and see if I can find some meaningful relatinships.
|