JEAVILY: Quantifying Market Entanglement

A high-dimensional signal processing engine for prediction markets. Jeavily maps hidden "entanglement" between seemingly unrelated global events using volatility Z-scores and entropy analysis.

Comment

Inspiration: The Physics of Finance

Most prediction market dashboards treat events as isolated islands. They show you that a candidate's odds are rising, but they fail to explain why. We hypothesized that global events—from Fed rates to pop culture—are "entangled" in ways that are invisible to the naked eye but detectable through high-dimensional signal processing.

We built JEAVILY to prove that prediction markets function less like casinos and more like a global neural network. We wanted to build the "Butterfly Index": a metric that quantifies how a flap of a wing in one market (e.g., Oil Prices) causes a hurricane in another (e.g., Tech Stocks).

How We Built It

Jeavily is not just a dashboard; it is a protocol. We leveraged Hex as our computational backbone to run a custom Python engine we call the EntropyEngine.

  1. Ingestion: We built a custom loader to fetch and normalize tick-level data from liquid prediction markets (Kalshi).
  2. The Math Core: We aligned disparate time-series data (fixing the "temporal mismatch" between hourly and daily markets) and ran a vectorized Pearson correlation matrix across the entire dataset.
  3. Shock Detection: We calculated the Rolling Volatility Z-Score for every market to detect "Panic Events" where $\sigma > 3$.

$$Z = \frac{x - \mu}{\sigma}$$

  1. Visualization: We used Hex's native integration with Plotly to render a "Dark Mode" Entanglement Map, allowing users to visually inspect the hidden correlations between markets.

Challenges Faced

The biggest hurdle was Temporal Alignment. Prediction markets are asynchronous; a "Climate Bill" market might trade 500 times an hour, while a "Mars Mission" market trades 10 times. Simple correlation fails here.

We had to engineer a resampling algorithm that "fills forward" the implied probabilities without introducing look-ahead bias, ensuring that our Entanglement Matrix represented true market sentiment at any given millisecond.

What We Learned

We discovered that volatility is contagious. When a "macro" market (like US Inflation) experiences a shock ($Z > 3$), it triggers immediate repricing in seemingly unrelated "micro" markets. Jeavily doesn't just show the odds; it visualizes the anxiety of the collective intelligence.

Built With

Share this project:

Updates