Prediction Market Analysis

Separating signal from noise in prediction markets using volume, volatility, and liquidity.

Comment

About the Project

Inspiration

Prediction markets have quietly become one of the fastest-moving sources of public belief. From elections and economic outcomes to sports and cultural events, market probabilities often update before polls, media narratives, or expert commentary.

What intrigued me wasn’t whether prediction markets are right, but when they are informative. Some markets feel like clear aggregations of collective intelligence, while others appear thin, reactive, or speculative. This project was inspired by the question:

How can we tell when a prediction market reflects real information versus noise?

What I Learned

Working through this project reshaped how I think about probabilistic signals:

  • Movement alone is misleading — large probability swings without volume often reflect noise rather than information.
  • Stability can be a feature, not a bug — markets that move slowly but attract consistent participation tend to form more reliable consensus.
  • Interpretability matters — a composite score is only useful if users understand why it behaves the way it does.

Most importantly, I learned how powerful it can be to combine analytics, semantic modeling, and AI into a single, explorable experience rather than a static dashboard.

How I Built It

I built this project entirely in Hex, using its notebook, semantic modeling, and app-building capabilities.

Data & Modeling

  • Ingested prediction market data including probabilities, volume, and liquidity proxies
  • Normalized time-series data to enable consistent analysis across market categories

Semantic Metrics

At the core of the app is a Trust Score, designed to capture how informative a market’s probability is:

[ \text{Trust Score} = 0.4 \cdot \text{Volume} + 0.3 \cdot (1 - \text{Volatility}) + 0.3 \cdot \text{Liquidity} ]

This score is defined once in Hex’s semantic layer and reused throughout the app for leaderboards, filters, and deep dives.

Exploration & Storytelling

  • Built interactive leaderboards for Top Movers and Most Trustworthy markets
  • Used a volatility–volume scatter to visually separate signal from noise
  • Added market-level deep dives showing probability and trading behavior over time
  • Layered in AI-assisted explanations to help interpret why a market moved

Challenges

Several challenges shaped the final design:

  • Liquidity is hard to measure — not all markets expose clean depth metrics, so I relied on normalized proxies.
  • Thin markets skew perception — small markets can show dramatic moves that look meaningful without context.
  • Balancing simplicity and rigor — the Trust Score needed to be expressive enough to matter, yet simple enough to explain clearly.

Hex made it possible to iterate quickly on these challenges by refining metrics, visuals, and narrative in a single workspace.

Reflection

This project reinforced that prediction markets are best understood not as crystal balls, but as information-aggregation systems with varying signal quality. By focusing on liquidity, stability, and participation — not just price movement — this app helps distinguish meaningful collective intelligence from short-term speculation.

Built With

Share this project:

Updates