One of my projects that I’m most proud of is my NFL professional football prediction system. I call it Zoltar.

The origins of Zoltar go back many years to my senior project when I was getting my BA in mathematics. Conventional wisdom at the time was that, “There’s too much variance in the system so prediction is not possible.” But I was stubborn and believed I could create a successful prediction system. I used classical statistical techniques and . . failed.

But I tried different approaches over the next few years. Finally, in a eureka moment, I realized that a system which used reinforcement learning (although I didn’t know the term at the time) could succeed. And it did.

Zoltar

Since that time there have been remarkable advances in all kinds of machine learning and deep neural techniques. I now believe that it’s possible for a team of about four people (three engineering types and one program manager type) to successfully predict all kinds of sports-related wagers.

The repeal of the PASPA (Professional and Amateur Sports Protection Act of 1992) law in 2018 (about a year ago as I write this) has opened the possibility of a tidal wave of new sports betting efforts. There is frantic activity by pro sports team, state legislatures, traditional casinos, and others, to create many different kinds of sports betting.

There is intense activity related to sports betting as a result of the Supreme Court repeal of the PASPA Law.

I’ve often thought about creating a startup company to do just that. I’ve dealt with venture capital people before and I’m positive I could get all the funding I need. Or, another approach would be to go to a large company (Facebook and Apple come to mind immediately) and propose a small, four-man team. Again, based on my experience, this would be easily accomplished.

But there’s a big problem. For me at least. When I predict NFL scores using Zoltar, I rarely bet actual money. For me the thrill is the intellectual challenge and the technology. I know where I picked up many of my personality traits but I have no idea where I picked up the notion that earning money has to be the result of providing some sort of useful value. If you win money by playing a slot machine or betting on a sports team, no value to anyone has been added. So it’s just wrong in some way (in my mind).

The sports book (betting area) at the Mirage Hotel in Las Vegas.

By the way, these ideas about sports betting are closely related to investing in the stock market. You can think of the stock market as a huge game. You invest in a stock and hope you will win when the price of the stock goes up. However, the big difference is that buying stock adds value to the overall economic system. So it’s not wrong (in my mind).

Finally, what about predicting the stock market? Conventional wisdom is, “There’s too much variance in the system so prediction is not possible.” I have some ideas about how to apply deep reinforcement learning to the problem of stock price prediction. And I’m quite certain that the stock market is beatable in a few narrow scenarios (but still worth tens of millions of dollars).

Gambling can be cool and sophisticated, but also rather sleazy at times.