cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 
JMP is taking Discovery online, April 16 and 18. Register today and join us for interactive sessions featuring popular presentation topics, networking, and discussions with the experts.
Choose Language Hide Translation Bar
Statistics as a 'superpower' for making fantasy football projections and predicting the NFL Draft

Do you love football -- the NFL Draft, fantasy football and football video games? Then, start following Joseph Bryan of Koalaty Statistics on Twitter!Do you love football -- the NFL Draft, fantasy football and football video games? Then, start following Joseph Bryan of Koalaty Statistics on Twitter!

For the past couple of years, I've been following Koalaty Statistics on Twitter, interested in the analyses and predictive modeling of football data tweeted there in threads. Sometimes the data that is explored and analyzed comes from video games, like Madden football from EA Sports, which releases its own ratings of players and teams. (I have a teenage son who loves both football and video games who teaches me about these things.)

With football starting up again soon, I DM'd the person behind the account, identified in the bio as an analytics specialist with a B.S. in statistics from the University of Georgia. It turns out Koalaty Statistics (@KoalatyStats) is Joseph Bryan, and he was kind enough to tell me a bit about himself and what he's up to.

Born and raised in Georgia, Joseph loves football – both college and NFL. While he had planned to study mechanical engineering at the University of Georgia, he switched his major after the professor for his basic statistics course taught concepts by applying statistics to football. Fast-forward a few years, and now this analytics specialist with a statistics undergrad degree has a following for what he calls his "nerdy regression-based take on football."

He got to work with data for the first time in a "real-world application" during an internship in quality engineering at Caterpillar. For the past year, he served as a quality engineer for Masterack in Georgia, where he helped introduce data analytics to the facility and led projects with estimated savings of more than $200,000 a year (!).

He got married early this summer and moved to Philadelphia with his wife, who is getting a PhD in economics at the University of Pennsylvania. And now he is on the hunt for a data analyst job in his new city. Meanwhile, he's been sharing analyses on Koalaty Statistics.

Tell us about Koalaty Statistics? What is it, and when/how/why did you start it?

OK, so a few years ago, I took a class called Applied Linear Regression, which was my favorite class at UGA. My professor, Mark Werner, was fantastic and truly made me love regression. That was a spring class, and I wanted to apply what I had learned to what I love (football). So, I started working on a regression to predict the NFL Draft. The draft is a huge NFL event that happens in April and one of my favorite things. I wanted to predict the draft before it happened or at least try. I created a regression, and it was OK, but nothing amazing. I collected all of the data manually (I didn’t have any of the skills I do now), and it took around 30 hours to get everything in a usable manner.

A little time passed, and my graduation approached, I thought, “I should start using my degree for football stuff... maybe people on Twitter will care.” I used my more refined statistical skills/methods and did the 2018 NFL Draft for running backs with *awesome* success. I predicted the draft with a .67 R^2 before it happened. I was super proud and started developing a small following. Especially, thanks to Peter Howard (@pahowdy) who was the first person to RT my work.

From there, I started applying my methods/skills to more than just the draft. I figured I needed a name, and my nickname in high school was Koala....so Koalaty Statistics was formed. I started working with regression in DFS (Daily Fantasy Sports), stuff like DraftKings and FanDuel, and my following got much larger. Right now, I am at around 1,803 followers, which is a ton for a nerdy regression-based take on football. When I first started, I never thought I would even pass 100 followers. It all started with the NFL Draft. But it has evolved into almost anything regression-based (within football). I do enjoy predicting DFS, which is essentially on a game-to-game basis.Toward the end of 2018, Joseph Bryan made this Bayesian model using JMP. It predicted whether a running back would be “good” or “bad.” Good = above the avg FP (fantasy points) scored & Bad = below the avg FP scored. The goal was to add another variable to a larger ensemble model that predicted weekly success. The X axis is FP scored in a single game.Toward the end of 2018, Joseph Bryan made this Bayesian model using JMP. It predicted whether a running back would be “good” or “bad.” Good = above the avg FP (fantasy points) scored & Bad = below the avg FP scored. The goal was to add another variable to a larger ensemble model that predicted weekly success. The X axis is FP scored in a single game.

How have football fans responded to Koalaty Statistics?

Usually, “awe” like, “how did he predict that a month before it happened?” Or when I predict a breakout player in a given week, people are pretty happy. Which is truly fun! The coolest moments are when people see what I do and are interested enough to DM me to learn more about the statistical methods.

What do you hope is the outcome of Koalaty Statistics?

The end goal is to help people learn statistics in a fun way. Every time someone asks me what I went to college for and I say "statistics," they *always* say, "Ewwww, that's hard/boring" (or some variant). Statistics is SO fun, interesting, and once you learn enough, it can be insanely useful.

Do you always win at fantasy football?

Haha, sometimes. My models are not perfect 100% of the time – no model is. I am still working on everything, and hopefully one day I can. I have had weeks where my model was almost perfect, but I have also had weeks where my model barely got anything correct.

What do you like best about statistics?

The whole reason I swapped majors in college was because I thought, “With statistics, I can predict the future,” and to a certain extent that is true. It’s like a superpower, and doesn’t everyone want a superpower?

How did you start using JMP? Why do you use it now?

I was first introduced to JMP in my applied linear regression course at UGA. We were taught how to use both R and JMP in that course. I still use both today, but mainly JMP for its fantastic visualization and neural network capability. JMP streamlines most regression tools, making it very easy to try multiple different approaches to a problem in a short amount of time. 

What else might people like to know about you?

When football isn't on, I enjoy watching NCAA softball and NBA basketball. I was in the Redcoat Marching Band at UGA where I played trombone. I am also an Eagle Scout. Lastly, during my senior year at UGA, I taught a coding class to kindergartners for an after-school program at a local school.

Last Modified: Apr 22, 2020 5:32 PM
Comments