Prediction markets like Polymarket and Kalshi often price the events differently and these inefficiencies create arbitrage opportunities but they are hard to detect manually and disappear quickly. Instead of just building a spread calculator, we wanted to build an system that understands markets and tells users exactly what to do.

  • Detects pricing discrepancies between equivalent contracts or markets
  • Calculates risk-free or low-risk arbitrage opportunities
  • Uses Google Gemini to analyze contract descriptions and context
  • Generates clear trade recommendations using Gemini API (what to buy/sell and why)

How we built it

  • Pulled live market data from Polymarket and Kalshi APIs
  • Built a matching system to identify equivalent events across platforms
  • Calculated probabilities and arbitrage spreads
  • Implemented a profit and risk evaluation model
  • Integrated Google Gemini API to:
    • Interpret contract descriptions
    • Ensure that there os equivalence between markets
    • Analyze sentiment and risk
    • Generate human-readable trade explanations
  • Built a simple dashboard to display spreads and AI recommendations

The system continuously scans markets and flags actionable opportunities.

-Challenges we ran into

  • Matching equivalent contracts with slightly different wording
  • Handling real-time API rate limits
  • Ensuring Gemini responses were fast enough for live trading insights
  • Normalizing probability formats across two different platforms

Accomplishments that we're proud of

  • Successfully detecting cross-platform arbitrage opportunities
  • Integrating Gemini in a meaningful way beyond simple text or summary generation
  • Building an AI layer that explains financial decisions clearly
  • Creating a working prototype during Hack Arizona HackDay

What we learned

  • Prediction markets are inefficient
  • AI reasoning significantly improves market interpretation
  • Contract wording matters more than expected
  • Clear AI-generated explanations increase user confidence

Built With

Share this project:

Updates