P-Pascal

Inspiration

The chaotic group chats between friend groups are one of the greatest things ever. But when your friend Dave claims he can get to sleep before 11 PM, get an A on an exam, or eat 50 nuggets in ten minutes, there’s no way to actually quantify how wrong he is.

What it does

P-Pascal is a live social prediction market that uses AI to set realistic starting odds for small groups, turning group chat arguments into tradable assets.

How we built it

We combined an LLM that generates Bayesian priors with the Logarithmic Market Scoring Rule (LMSR) to ensure continuous liquidity and mathematically sound pricing.

Challenges we ran into

Finding the venue on day 1...

Accomplishments that we're proud of

We successfully shipped a fully functional platform during the hackathon, got real students to place bets, and produced a technical whitepaper detailing our liquidity calibration, mathematical reasoning, and everything else. It's rewarding to see people have fun with the platform.

What we learned

We discovered that "Thin Markets" (2–15 people) require a carefully tuned liquidity parameter (b) to make prices responsive enough to feel fun while maintaining market stability.

What's next for P-Pascal

We’re looking into dynamic liquidity scaling so the market doesn't break when Alvin’s entire network of friends decides to bet against him.

Built With

• git