I am not sure how to introduce this subject except to say that I will be glossing over a lot -- partially because I want to keep my work somewhat private and also because an in-depth discussion would get deathly boring.
In college, I was fascinated by the potential to apply artificial intelligence algorithms to the stock market and made it one of my senior projects. The first incarnation was an expert system based on one of Warren Buffet's books. This ended up being nothing special since one could easily get better results with a pen and paper and Yahoo Finance's stock search tool. But you know, you gotta start somewhere.
The biggest problem with using expert systems is that the subject of finance is too broadly studied and profitable. Expert systems are geared towards esoteric subjects that few know anything about and even fewer would want to learn.
I spent several years believing that there was no advantage to be had in the markets until I read the book Trend Following by Michael Covel a few years ago. More interesting than the interviews with the experts to me was just the basic concept: that trends have always existed and will always exist as long as humans are involved. The efficient market hypothesis is wrong or else how would these guys have made money? Furthermore, the Incompleteness Theorem would imply (to me at least) that the efficient market hypothesis cannot be true, but I digress....
Since then I have developed an evolutionary algorithm based hedge fund program that does quite well indeed....live simulations of my latest version have yielded a 17.6% return since April 15. The problem there is that the initial investment would ideally be very high and the potential returns are quite low.
That is when compared with the potential that the same formula could be used to develop a sports betting algorithm, which I have also been developing over the last two years. So far I have only done well in college football during the 2013 season, but there have been a lot of improvements made since then. The current version is being applied to baseball betting, and has done quite well indeed. I have broken even over thousands of bets so far this season as improvements have been made and bugs fixed, which would indicate a slight edge over random. The latest version is running at a very good clip though.
So has anyone else pursued the idea of a game-playing/trend-following algorithm for sports betting? Results good or bad? Pitfalls? Sticking points?
Now for something hopefully less boring....today's picks:
Toronto, Boston, Oakland, Chicago White Sox, Kansas City & UNDER, LA Angels, Baltimore, St. Louis, Arizona & UNDER, Washington, Atlanta, NY Mets, Pittsburgh, LA Dodgers, Miami & UNDER