Show Your Work - Machine Learning for Football
Updated: Oct 16, 2019
This summer's project is coming to shape, and I hope you are interested, a book on how to build Machine Learning models for NFL games: Machine Learning for Football A Guide to Building Sports Models.
I am currently reading Show Your Work by Austin Kleon, it is his follow-up to one of my favorite books, Steal Like an Artist, which I discuss here. As he writes on his site, "If Steal was a book about stealing influence from others, Show is about influencing others by letting them steal from you." With that in mind, this post is about the decisions leading me to write this book, who's it for, and what's it for.
Starting with why I am writing this,
1. Football Season Ended, and my Picks were Awful...
Well maybe I should say "My Picks were Average", as in on par with a blind monkey flipping a coin (or throwing darts if you are a fan of indexing-style of investing). To be clear, there is a reason the lights stay on in Vegas, it is NOT EASY to beat the line. However, my process of listening to podcasts and reading articles throughout the week led to poor outcomes.
The small sample size and randomness of sports games easily can contribute to a poor season from a good process (a good process doesn't guarantee a good outcome), but bottom-line, I was not happy with my process as it lead me to find one or two interesting stats, look into a pick based of those few insights, and (here is where my error lies) then I'd find myself shaping the rest of the decision to fit that pick (i.e. selecting only what already fit my narrative).
2. The Book I Want to Read
Knowing there are errors in my process, what is the best way to let the data guide me?
By building a model that picks the right team every time! Of course!
Letting an algorithm solve your problems can be foolish, especially when so much randomness and intangibles go into every game. However, we can build models to point us in the right direction and to make predictions that we can dive deeper into it before making the pick.
Big Hairy Audacious Goal - Build a model that picks 60%
If a model truly picked 60%, it would easily make it one of the best programs in the world of sports betting. Even though that goal is like solving an industry-wide issue with one project (not likely) it is a target that I am aiming for.
A Model that Predicts 60%? Bring it On.
If you are a fan of author and speaker Seth Godin, I recommend you look into the altMBA. If you have never heard of him, then I recommend you look into the altMBA. It is an month long, intense online workshop focusing on leadership and change management. I did altMBA27 in February 2019. Part of the work within that program helped me define my goals, understand how I can make an impact on the world (personally and professionally), and in essence opened my eyes to seeing how decisions and change really happens.
It also helped me to see many amazing people (your peers are in many ways the best part of the program) and what they are capable of.
Write a book?
Build a model?
As the resistance (the thing inside of you that tells you should avoid this, you are not good enough, who are you to try that?) screams at me to stop, I know my altMBA colleagues (and my supportive wife) have my back.
Once you see your potential, you cannot unsee it.
Who's it For?
It's for football fans that are interested in building mathematical models to help them better predict and understand the game.
What's it For?
To allow these fans, who are interested in building these models, to walk with me as I learn and build models for myself and to give them confidence and the tools they need to do it themselves.
When will it Ship?
It will be available on Amazon on July 25, 2019.
Thank you for reading.