Team Matrix (Overview)
Team Matrix (Selected)
Play-By-Play (Top)
Predictor (Bottom)
Predictor (Top)
Play-By-Play (Bottom)

Inspiration

Sports fans often turn to ESPN or similar sites for NFL data, but the information is fragmented, overwhelming, and purely descriptive. There’s no built-in way to compare two teams head-to-head with predictions or confidence scoring. We wanted to make a tool that simplifies this experience, one platform where fans can get insights, predictions, and interactive simulations without digging through endless stats pages.

What it does

Our project is a comprehensive NFL analysis and prediction dashboard for the 2025 season. It provides:

AI-powered predictions: Compare any two teams and see predicted winners, final scores, confidence levels, and key matchup advantages.
Team analytics dashboard: Explore detailed statistics for all 32 teams, including points per game, team record, and yards per game.
Play-by-play simulation: Watch 2025 NFL games unfold with real player data, ball position, and interactive playback controls.
Real-time integration: Data stays fresh with automated espn api and nflreadpy updates and caching.

Essentially, it transforms raw NFL stats into clear, predictive insights and engaging visualizations.

How we built it

Backend: Flask server handles espn api and nflreadpy calls, caching, and data cleaning.
Prediction model: A machine learning system trained on historical NFL data from 2021-2023, tested against data from 2024, and now predicting all current matchups between teams weighted by offensive strength, defensive performance, head-to-head performance, home field advantage, recent streaks and many more. Outputs include win probabilities, score predictions, confidence scores, and reasoning for predictions.
Frontend: Built with HTML, CSS, and JavaScript for responsive design.
Play-by-play engine: Processes espn api and nflreadpy play-by-play data into a dynamic simulator that visualizes each snap, down, and yard gained in real time.

Challenges we ran into

Training the prediction model: Balancing statistical rigor with hackathon time constraints was tough. We had to carefully weight historical data while still delivering interpretable results.
Data parsing: espn api and nflreadpy provides a lot of raw play-by-play detail, which required custom parsing and error handling to structure correctly.
Performance trade-offs: Ensuring real-time updates while caching effectively to avoid repeated heavy data pulls.
UI complexity: Combining analytics, predictions, and simulations into one interface without overwhelming the user.

Accomplishments that we're proud of

Building a working ML-powered prediction system that produces realistic outcomes with confidence scores.
Developing a full 2025 analytics dashboard that covers all 32 NFL teams in detail.
Creating an interactive play-by-play simulator that enhances fan engagement with real game data.
Delivering a clean, modern user interface that simplifies complex stats for casual fans.

What we learned

How to apply ML techniques to live sports data and make predictions interpretable for fans.
How to structure and optimize espn api and nflreadpy play-by-play data for analysis and visualization.
The importance of presenting advanced analytics in a way that feels fun and approachable, not overwhelming.
How collaboration and rapid prototyping helped us achieve a full end-to-end product in a short timeframe.

What’s next for Snap Stats

Expanding beyond the NFL to other leagues (NBA, MLB, NCAA).
Training the ML model with deeper datasets (player-level stats, injury reports, weather factors).
Adding fantasy football integration so users can project player impacts on outcomes.
Building a mobile app for on-the-go predictions and simulations.
Enabling social sharing features, where users can compare predictions with friends or communities.