High School Research Student Vibhu Uses Sports Analytics to Help Football Coaches
Vibhu is a senior from San Jose, California who researched sports analytics with the help of his Polygence mentor, Ben. Using a 10-year data set that included 2,560 football games, Vibhu created a user-friendly tool named, Conversion Tool, that aids coaches in making decisions when it comes to fourth downs and two-point conversions. He also wrote a Medium article, which you can find here and is currently working on a new website that will allow users to rank players based on their statistics. You can read more about Vibhu’s Polygence experience in the interview below.
Why did you decide to do Polygence?
I heard about it because of my counselor. We wanted to figure out something to do over the summer that wouldn’t be super boring and that I could enjoy doing. I also got to choose my own project, which was cool because I always wanted to do something in sports analytics. I was on the fence about computer science or data science when I got started, but I think Polygence definitely inspired me to do something in those fields now in college.
More specifically, I studied how to help football coaches make decisions when it comes to fourth downs and two-point conversions based on previous data from games. I think I was also inspired by knowing that the Baltimore Ravens actually use sports analytics on fourth downs.
I was on the fence about computer science or data science when I got started, but I think Polygence definitely inspired me to do something in those fields now in college.
Oh, cool! How do you even do that?
Ben, my mentor, suggested I use the programming language, R, and we took the play-by-play data from the past — I want to say from 2009 to 2019. We wanted to find the probability of a situation happening based on that previous data and given a user’s inputs. Then, you can find the probability that a team wins if that situation occurs. For example, if you get the fourth down, you have to find the probability that you win on that and if you don't get the fourth down, you find the probability that you win on that too. So, you use a statistics-like perspective to find an average winning probability for these different decisions.
How did you come up with this idea? In some ways it seems like a simple idea, to help a coach make decisions, but what you just described is pretty complex.
I went into Polygence saying I wanted to do something on sports analytics. Then, I'm pretty sure one of the project ideas under Ben’s profile on Polygence’s website mentioned helping a coach make a decision. So, I narrowed that down a little bit because I love football. It's probably one of my biggest passions. So, I decided fourth down and two-point conversions would be cool. I also wrote an article about my project on Medium which describes the use of analytics and my tool in football. It also details how I created my tool and the statistical approach I took when coming up with these probabilities. At the end, I entered my research in a competition to get displayed in the MIT Sloan Sports Analytics Conference.
What was something that surprised you either while doing your project or looking through the results of your project?
Well, we created this tool, which I used to loop through a ton of situations — like game time for down decisions — but then I wanted to validate it somehow. I wanted to see when teams made the wrong decision on 4th down, according to my tool. For instance, when my tool said to go for a 4th down, but the team actually punted it; or vice versa. So, I filtered the data set to only fourth downs from every team from the past two years and then, put each of these situations in the tool and counted each time where a team made a wrong decision according to analytics. When the results came back, I counted those for each team and saw that the Baltimore Ravens had the least amount of wrong decisions made for fourth downs. I was super happy about that because I knew from before that Baltimore actually uses sports analytics on fourth downs, and so that kind of just validated my tool.
What were your sessions with Ben like?
I usually came in with any questions or errors I had, and he would walk me through how to fix it or help me to think through a solution. After that, we wanted to use our time more usefully, so we wouldn’t spend any time coding, but we did make a task list on what to do to get farther in my project and how to go about it. And so, if I had questions later, I could have a general idea of how I would code this.
We would also plan my work for the next week by pseudo coding, which is basically like planning how I would code something or how I would create an algorithm in english, rather than actually coding it in R. This is a super useful technique because it allows us to brainstorm easily and to get a general idea of how I would code something without actually having to code it and taking up too much time during the session.
How was Polygence different from other learning environments you've been in, in the past?
It's interactive and not like a lecture session, which would be pretty boring. I’ve never had a teacher or mentor in a one-on-one type of setting. I've only had group classes, so this was super different since Ben focused solely on me, rather than an entire class. Ben was probably one of the most fun teachers I've had. I don't even really consider him as a teacher. He’s more so a mentor or a friend. He’s just super nice and friendly and knowledgeable. We're both doing different things and talking the whole time too. He's super entertaining!
Ben was probably one of the most fun teachers I've had. I don't even really consider him as a teacher. He’s more so a mentor or a friend.
Do you have plans for your next sports project?
Yeah, in R, I can do analytics on each player based on their available data and their stats. So, now I want to start creating a model in which I can work with a data set with tons of different players and their stats. Later, I want to be able to browse at my will and create predictive models based on this dataset for each player. And then, I want to turn these stats and models for each player into a website just to make it more user-friendly, so anyone can search a player and find their stats and predictions. This could even turn into rankings, which can be used for Fantasy Football.
What advice would you give to a student about to start a Polygence project?
I would say choose a project that you're super interested in, so it'll be fun instead of just something that you dread to do every day. I’ve said before that I love football and now, I look forward to coding every day. I actually have the drive to do it because my project applied my coding skills to what I am knowledgeable about and I'm interested in.