Hi all,
I am an alumni of the first round of Jeremy Howard’s class, and I wanted to share what I’ve been up to.
I used deep learning to build a MLP which ultimately allowed me to put together a system for gambling on boxing. Is this what Jeremy had in mind for his illustrious alumni? Ummm…maybe not exactly…but I’m happy to say I’ve gotten amazing results.
The TL;DR is this:
“The model, with a decision threshold of 0.90, was given the opportunity to place 679 bets. It chose to place thirteen bets, winning all but one. With an initial hypothetical investment of $1,300 ($100 per bet), it won $292, which represents a ROI of 22.5%.”
Who needs Bitcoin when there’s this opportunity for speculation?
I would be thrilled if anyone wanted to read my blog posts about it. And even more thrilled if anyone would comment and give me accolades/spankings for mistakes.
It’s a set of 3 posts:
http://www.mattobrien.me/deep-learning-for-sport-wagering-part-1-of-3.html
Also, I wanted to say that I’m going to be coming back to retake the course again. The first time around was so fun that I couldn’t resist coming back to do it all again. This time in PyTorch? Let’s do this!
Matt