This all started when I used to look at the NY Times computer rankings of football teams they would publish in the paper, since discontinued. I was curious if you assembled that data whether you might develop an edge against the line. I doubted it, but was curious nonetheless. As I crunched the numbers, I developed new ideas for different variables, tested them, developed new ones, etc. Gradually I developed a model that seemed to work. As I was doing this as a hobby, my data sets were rather intermittent, often starting midway through a season, and often I had trouble finding historical data for the variables I was pulling down from the web.

As this model kept performing, I naturally wondered: hey, I tried one sport, it looks like it is working reasonably well, maybe others could work. Again, I was skeptical but curious. When I left my prior job after 7 years, I decided to take some time off and thought it would be a good time to work (after some travel to Europe and South America) on the models in earnest. Or, at least test some hypotheses to see if further research was warranted. Despite my skepticism, part of the pull is that I love working on this models. There is something oddly compelling about developing a hypothesis, assembling the data to test it, analyze the results and assess what to do next. I can work on this stuff seemingly forever. Despite the success of the NFL model and my genuine love of crunching the numbers, my analysis into other sports has not yielded the results that encourage me to go deeper. I have ideas, possibly very good ones, but the economic realities of life limit how much time I can spending working on this stuff. But it is fun.

So, the pages in this section highlight some of my methodology in developing the models, variables I tested and the results. With the NFL model, because it still seems to be working, I show mainly the results of it rather anything too revealing about the inputs.