Ongoing side project with a few friends
Fantasy football is what initially got me interested in data science when I was in 6th grade. I remember copying and pasting all sorts of data into Excel to try and build a model that could help me gain an edge. I've continued to refine my strategy and analytics over the years, and in 2017 a few friends and I decided to join forces to try to create the best fantasy football predictions available. Last season we beat ESPN by 23% and tied the historically best rankings (4for4).
My family is very competitive when it comes to fantasy football. If you don't put in for pickups and play a full lineup every week, you'll likely be kicked out next season. My dad, grandpa, and uncle basically live with premium football talk radio playing all the time so it's hard to gain an edge in our league. We actually have a pretty cool trophy that the champion gets to house for a year. It is my pursuit of this trophy that drove me to start Gridiron AI with my friends.
Most of what we do is scrape and store a ton of data. It takes all three of us (since it's a side project) to keep the data collection running during the season. We take this data and feed it to machine learning algorithms that learn to predict player stats in the next game. It's in this way that we can then convert those stats to PPR, standard, or custom scoring for our users. We put our rankings online for a simple, $5 subscription to help us cover the costs of our data collection and modeling. We still don't break even, but we are hoping next year we will! Either way, we all have a lot of fun with it and having access to all the data we've collected is worth all the work. We are all obsessed with fantasy football and would take any excuse to spend time giving our teams an edge.