Published on Dec 12, 2019
My family loves fantasy football. We’ve had a highly competitive 12-team league since the early 90s and strictly enforce that everyone must submit lineups every week, put in for pickups, and not cheat (combining teams, dropping good players, etc.). If you break the rules, you get kicked out.
I’ve had my own team since I was ten years old—about sixteen years. I’ve watched football every Sunday for my entire life and have always tried hard at fantasy football, but I’ve never made it to the Super Bowl (or won).
Some of this is based on bad luck. I’ve had some good years where I finished with the highest points during the regular season, but lost in the playoffs the first or second round.
This year, though, my wife and I are facing off in our Super Bowl. It’s only her second year playing, so she’s of coursing giving me a lot of shit.
The reason I’m bringing this family saga up is that about 2.5 years ago two strangers whom I met online and I decided to build a model for fantasy football and pick’em predictions. We called our model Gridiron AI.
We knew our model was getting pretty good last year when we finished with the most accurate rankings across the year among major fantasy football websites.
So this year, I drafted based on our rankings, chose lineups based on our rankings, and put in for pickups based on our rankings.
They led me to draft Austin Ekler, Cooper Kupp, DJ Moore, and Julian Edleman higher than most sites had them ranked. These are the best players on my team.
I went against our rankings on two major occasions drafting Cam Newton and LeVeon Bell. Those have been my two worst picks.
When Cam went down, our model had Stafford ranked as the highest available QB in our league. I picked him up and he was great for several weeks… Until he too got hurt.
Then our model told me to pickup Ryan Tannahill, and that decision might help win me our super bowl (he’s been killing it).
It’s not just my team that’s used our Gridiron AI rankings. My wife also used it throughout the year to draft, choose pickups, and set lineups.
We are the only two teams in the league who used the Gridiron AI rankings, and we’re both in the Super Bowl.
I know a lot of luck was still involved and maybe there isn’t enough data yet to call it statistically significant. But it’s still pretty cool.
Gridiron AI has been my favorite side project for the last two years. It’s something me and my two friends work on few hours per week for ~25 weeks per year. We don’t plan to ever make a lot of money off it. Our goal is for the website to one day offset our costs of collecting/storing data and computing power for training our models.
Nonetheless, it’s something I derive a lot of pleasure and value from.
I think this is an important lesson to remember. You can have side projects and hobbies that won’t make you rich.
Money is important in life because it gives you access to opportunities. I think too often we forget to take hold of some of those opportunities.
Gridiron AI is a great example of this. I put a lot more money into it every year that I’ll probably ever get out. For me, the experience is worth the expense. Through this side project I’ve made two good friends I probably wouldn’t have otherwise, I’ve taught myself machine learning techniques that I’m now using in my real job, I might finally win a fantasy football Super Bowl, and I’ve had a lot of fun in the process.
Here’s our Super Bowl matchup. Since all of the $600 betting pool goes to 1st and 2nd place in our league, we’re playing solely for glory (the money all comes back to the same house)!