Inspiration

We wanted to recreate the thrill of watching sports with friends by letting people predict the future in real time. Instead of static fantasy or pregame bets, what if you could bet on the next play as it happens?

What it does

Game Pulse transforms live sports into an interactive, second-screen experience. We stream sporting event transcripts in real time, analyze the flow of the game, and use an LLM to generate prediction questions instantly. Players join lobbies, answer these live bets, and climb leaderboards based on their accuracy and speed.

How we built it

  • Live Data: We simulated a real-time transcript feed of game events (plays, stats, outcomes).
  • LLM Integration: Each event triggers a prompt to an LLM, which generates a concise Yes/No question (e.g., Will the next play be a pass?).
  • App & Backend: The mobile app was built in React Native + Expo, while Supabase handled authentication, lobbies, player stats, and live updates.
  • Leaderboard Engine: Player bets are scored instantly and displayed in a dynamic leaderboard, updated after every question.

Challenges we ran into

  • Streaming and syncing real-time data across all devices.
  • Getting the LLM to generate reliable and fair questions from raw game transcripts.
  • Balancing latency so users get questions and results without delays.

Accomplishments that we're proud of

  • Built an end-to-end system: live data → LLM → interactive betting → leaderboard.
  • Made real-time dependent bets actually work in a hackathon timeframe.
  • Created an addictive, social experience that feels natural during a game.

What we learned

  • How to design LLM prompts to turn raw transcripts into structured questions.
  • How to manage real-time multiplayer synchronization under time pressure.
  • The power of combining sports data, AI, and social competition into one flow.

What's next for Game Pulse

  • Expanding to multiple sports and refining the LLM prompts for each.
  • Adding a marketplace where players can buy/sell sports-themed items with in-game winnings.
  • Scaling to handle hundreds of simultaneous lobbies with minimal latency.
  • Exploring integrations with official sports APIs for richer data streams.

Built With

  • gpt4.0
  • openai
  • python
  • reactnative
  • sportsdataioapi
  • supabase
Share this project:

Updates