Sports Analytics as a Student in College
How to apply what you’re learning in class to real sports problems
Tej Seth

Introduction

ABOUT ME
Tej Seth
  • Graduated from the University of Michigan
  • Majored in Information Analysis and minored in Applied Statistics
  • Online at @tejfbanalytics
  • Previous experience: Michigan Football, SumerSports, ESPN Bet
  • Current role: Data Scientist at DL Trading

Tej Seth slide image

My Sports Analytics Journey

How It Started

From baseball cards to football data

  • As a kid, I always looked at the stats on the back of baseball cards
  • I was a huge sports fan, especially Detroit teams

High school

  • Copied and pasted data from Pro Football Reference
  • Built scatter plots in Google Sheets (would not recommend)

My Sports Analytics Journey

Turning Point

College → Where things got serious

  • Joined the Michigan Football Analytics Society (MFANS)
  • Took my first coding class (C++)
  • Started to understand how sports analytics can be applied

Teaching Myself How to Code

Early Learning

Trying to recreate what I saw online

  • Started coding freshman year using R
  • Tried to recreate graphs from @benbbaldwin
  • Spent weeks debugging and figuring things out

Then everything clicked

  • Started to understand how to put together sections of code

The moment everything changed

# install.packages("nflfastR")
library(nflfastR)
library(tidyverse)

pbp <- load_pbp(2019)

pbp %>%
  select(posteam, defteam, down, 
         ydstogo, yards_gained) %>%
  head()

The Next Steps

The Next Steps

  • Started posting graphs and analysis on Twitter
  • Received a lot of constructive criticism (and some criticism)
  • Started stringing together multiple graphs for long-form articles
  • Enjoyed combining data and analysis in one central area

Data Viz Over Time (Bad Example)

Data Viz Over Time (Good Example)

Getting Your Name Out There

Getting Your Name Out There

  • Posting articles on a website and graphs on Twitter gets attention
  • Especially when you can reply with a graph you made to a bigger account
  • Tweets are fine, but making connections with people in the industry is one of the best things to do
    • LinkedIn
    • Email / Direct Messages
  • Bottom line: doing good public work will get you noticed

My Journey to Pro Football Focus

My Journey to Pro Football Focus

  1. Got in touch with a current intern there through Twitter DMs
  2. He gave me Eric Eager’s email (the Director of Research & Development)
  3. Eric had seen my tweets and articles and knew I liked football and data
  4. Applied to the link that he sent me and told him I applied
  5. Had an interview with him plus a screening with HR

Balancing School and Sports Analytics

School First, Sports Second

  • School usually came first because there was a timeline
  • Having weekend projects to look forward to made schoolwork go by faster
  • Doing sports analytics doesn’t feel like work because it’s sports

A Process for Generating New Ideas

A Process for Generating New Ideas

  • Watching a lot of football helps conjure up ideas
  • Look into other sports and see how they are measuring performance
  • Start with early exploratory data analysis (EDA) on the topic
  • Make a simple model to see if the process is there
  • Go to an advanced model if needed
library(nflfastR)
library(tidyverse)
library(ranger)

pbp <- load_pbp(2022)

pbp_pass <- pbp %>%
  filter(pass == 1, !is.na(air_yards))

lm_air_yards <- lm(
  air_yards ~ as.factor(down) + ydstogo + yardline_100,
  data = pbp_pass
)

rf_air_yards <- ranger(
  air_yards ~ down + ydstogo + yardline_100,
  data = pbp_pass
)

Resources & Competitions for Sports Analytics Students

Resources & Competitions

Tutorials

Competitions

Experience

SumerSports
Data Scientist

  • Consumer-facing analytics content
  • Fantasy tools: projections, optimizer, waivers

PENN Interactive
Data Scientist

  • Built NFL simulation engine
  • Developed play-level outcome models

DL Trading
Data Scientist

  • Built predictive trading models
  • Feature engineering and signal generation

The Future of Sports Analytics

Where Things are Headed

  • It will be run by students like you guys
  • Even if you do not decide to go into sports analytics, being able to use tools like R and/or Python to analyze, visualize, and model data is an insanely useful skill to have
  • Analytics in the pros:
    • MLB had its “Moneyball” takeover more than a decade ago
    • NBA just had its midrange revolution
    • NFL is next as teams start to use tracking data to their advantage

Conclusion

Conclusion

  • Always feel free to reach out to me at:
  • Thank you so much for listening to my presentation
  • Feel free to ask any questions in the Zoom chat or over the mic

“In the midst of chaos, there is also opportunity.” - Sun Tzu