Let’s have some fun with data

I’m Jake Davis. I lead data science, product, and strategy teams for a living, and I like building statistical models and performing analyses for fun.1

Jake Davis Analytics is a result of my hobby needing more room to breathe (and a forcing function for me to take things over the finish line). It’s a sports analytics publication. It’s mostly football, plus whatever else has caught my attention that week. The posts are (mostly) long-form and built from real analytical models, and I have a soft spot for questions nobody really needed answered.

What you’ll find here

Jake Davis Analytics seeks to explore questions - both good and random - with real rigor. One of my biggest struggles professionally and personally is the assumptions of simplicity that are often placed on surface level questions, and how that lack of depth leads to sub-optimal decision-making and flat out wrong conclusions. So you’ll find even the most trivial questions given the full analytic treatment. You’ll see the obvious ways these questions can be answered, where those hold up and where they break, and then a more formal or innovative approach. As an example, see Your record doesn’t know you like it thinks it knows you, on how little a 17-game season can tell you about a team.

The Interactives

One of the primary reasons I started Jake Davis Analytics was to force myself to finally build the interactive sites that had been swirling around in my head.

The first is for LOWFI - a game-level NFL forecasting system. It rates every team, projects every game, and simulates the rest of the season 10,000 times. It updates itself every Tuesday at lowfi.jakedavisanalytics.com.

More models and interactives are coming!

How I work

I mostly program in R. I tend towards traditional statistics where they do the job and machine learning whenever they don’t. I prefer practical over fancy. I try to write as plainly as possible, and I’ll generally leave the math and details in the footnotes. I publish as much of the code and data as I can (or remember to) on GitHub, so you can check my work or do something better with it.

Get involved

I love a rigorous answer to a slightly dumb question. If you’ve got one that deserves a few thousand words and a few thousand lines of code, reply to any post or leave a comment.

-Jake