I’ve built Aurca.ai, a web app to help people betting on prediction markets spot mispricing opportunities. It analyzes contracts using statistical models and machine learning by evaluating odds (from data) and uncover where the market might be off. I’ve been tinkering with this for a while, combining my love for data science and financial markets, and it’s finally ready for others to try.

You can jump in and check it out at aurca.ai/dashboard—no signup needed to play around with the core features. Just pick a prediction market contract (it supports Kalshi for now, with more to be added later), and it’ll show you the model results. My approach uses historical data, probabilistic models, and machine learning to predict events and evaluate their occurrence probability and compare that with the market price.

It’s early days, so expect rough edges—I’m here for feedback! What works? What’s confusing? Any features you’d want? I’ll be around to answer questions and tweak things based on what you think.

P.S. All Models are trained using Bayesian methods, and the tool used to train is Numpyro on JAX.