BrandBull

BrandBull predicts how sports sponsorships impact stock prices. Using AI, we analyze game results and sponsor ties to forecast market moves—giving traders a unique edge in financial decision-making.

Comment

🏆 BrandBull: Making Corporations Sponsor YOU!

🔥 Inspiration

The inspiration for this project came from the Taylor Swift Happiness Predictor project at Hacklytics 2024. That team won the Most Creative Hack award by coming up with a unique alternative indicator for a highly researched asset.

This got us thinking: Corporations pour billions of dollars into advertising and sponsorships—what tangible returns do they actually see?

We decided to take a deep dive into this question, and what we found was fascinating...

⚡ What It Does

BrandBull is an AI-powered tool that predicts how a company's sponsored sports team's performance impacts its stock price.

🔹 The neural network takes in:

  • A stock ticker associated with a company sponsoring a sports team
  • The outcome of a game played by the team
  • The sport that was played

🔹 It returns:

  • The predicted percent change of the stock on the next business day

While our model doesn’t claim that a single game can massively influence stock prices (since large-cap companies have many external factors affecting them), it does prove a measurable and predictable impact—highlighting a key principle of interdependence in economics.

🏗️ How We Built It

We followed a structured approach to tackle this problem:

🔍 Step 1: Data Collection

  • We built a web crawler/scraper to search for thousands of websites related to company sponsorships, sports sponsorships, and team sponsorships.
  • We extracted raw text and passed it through an LLM API to convert the data into structured JSON format.
  • Using an API, we mapped company names to stock tickers.

🛠 Step 2: Data Processing & Engineering

  • Isolated the time range each sponsorship lasted.
  • Gathered all games played by the sponsored teams during that period.
  • Extracted relevant data: game performance, sport, game date, and corresponding stock prices.

🤖 Step 3: Neural Network Training

  • Trained a deep learning model on the cleaned dataset.
  • Achieved validation loss < 0.01, MSE < 0.01, and MAE < 0.5—indicating a strong predictor without overfitting.

🎨 Step 4: Visualization & Deployment

  • Built a front-end visualization & stock pick tool using Streamlit.
  • Users can input a stock ticker and see how game performance impacts its price.

🚧 Challenges We Faced

✅ Data Complexity: Since the data came from highly heterogeneous sources, cleaning and organizing it into a structured format was extremely challenging.
✅ Automating Data Collection: Extracting sponsorship details in a scalable way required careful prompt engineering with LLM-based extraction.
✅ Ensuring Model Accuracy: Balancing a model that generalizes well while keeping its predictive power strong was key.

🏅 Accomplishments We’re Proud Of

🎯 Successfully consolidating a massive dataset from various sources into a cohesive, structured format.
🎯 Using generative AI in an assistive manner—rather than over-relying on it for core decision-making.
🎯 Creating an efficient, resource-minimal predictive model that extracts meaningful insights about stock price behavior.

📚 What We Learned

This project was a crash course in end-to-end data science:

  • 🔹 Data Acquisition: Building web scrapers and utilizing LLM-powered extraction.
  • 🔹 Data Cleaning & Engineering: Filtering noise and handling unstructured data.
  • 🔹 Machine Learning: Developing an AI model that effectively predicts stock price movement.
  • 🔹 Visualization & UX: Building a Streamlit-powered front-end for easy interaction.

And most importantly... we learned that sports sponsorships do impact stock prices in a real and predictable way.

🚀 What’s Next for BrandBull?

We’ve laid the foundation for a powerful data analysis tool—but there’s so much more we can do:

🔹 Scale Up Data Collection: Automate web crawling & NLP-based extraction for millions of queries.
🔹 Expand Model Coverage: Incorporate additional factors like earnings reports, social media sentiment, and macroeconomic trends.
🔹 Deploy as a Full-Fledged Product: Enhance user experience and create a real-time stock impact tracker.

💡 BrandBull is just getting started. The future of AI-driven financial insights is here.

Built With

Share this project:

Updates