LinkedIn Recruiter

A database collected to explore how media attention follows professional female athletes and whether game-play statistics uphold these narratives.

Build a database through scraping sports media to explore and analyze what language and narratives follow different WNBA players and teams over the past two seasons as the league has grown exponentially in popularity and media coverage (2024-2025). Explore game and season average statistics to discover if the loudest stories are justified with statistical evidence or if the data tells a different story.

Uncover and uplift largely untold stories, less covered players, and teams that fly under the radar.

I scraped articles from 8 news sources’ WNBA pages (see below), and after initial analysis of the headlines and summaries I had gathered, I decided to focus on the theme of physicality and explore how I could contrast additional quantitative data with the media language I had collected.

I then scraped Fox Sports’ Misc Stats page for technical and flagrant foul aggregates throughout the season, and scraped wnba.com to get the game schedule for each season. I used this to scrape each game’s play-by-play data, gathering video source URLs for each play that involved a foul and collapsing all of this data into a Pandas dataframe with lookups by player, team, and game.

I worked on categorizing articles by teams and players by building entity lookups using players’ last names, weighing some players above others depending on which teams were first detected and filtering out non-WNBA overlap names (e. g. people like Serena Williams who show up on WNBA-related articles because they are investors but shouldn't match to articles about Courtney Williams or Elizabeth Williams).  

I decided to normalize personal fouls, technical fouls, flagrant fouls, and fouls drawn over minutes played to get an idea of how much players were fouling or getting fouled related to how much they're actually on the court as opposed to the sometimes inflated numbers we see reported out of this context. The final step was merging the foul data with the articles data, which I linked using player/game/team ids.

I built a website to display both the qualitative and quantitative data I collected using React, Next.js, and some beginning sentiment/language analysis Javascript libraries like Wink and Rita. It functions as an exploratory interface for the database, organized in pyramid-like structure with the highest number of headlines at the bottom so that less-featured player/game stories are uplifted by being displayed on the top of the page.

The main goal of this visualization was to contrast the qualitative data like headlines and videos with the normalized foul data to give the viewer a holistic understanding of the topic of physicality in the WNBA.

This project is very much still a work in progress - now that the data has been gathered, cleaned, and organized, some next steps for phase two will include:

Increased language analysis and categorization 

Pulling out language sentiment themes manually then using Claude & Gemini to iteratively categorize remaining articles and identify more themes

Python, Pandas, Beautiful Soup, Playwright, React, Next.js

WNBA Storylines

Case Study Coming Soon