PredictionBubbles: 63% price may not equal 63% odds

PredictionBubbles launched a live dashboard in mid-August 2026 that aggregates markets from Polymarket and Kalshi into a single bubble view. Each tradable question appears as a bubble displaying probability, recent price movement and trading volume.

Bubbles can be sized by open interest, 24-hour volume or an activity ratio and filtered by category, platform, expiry, probability and open interest. The layout resembles stock and crypto heat maps and provides a discovery layer for markets that typically require opening individual pages to see detail.

Polymarket provides public market-discovery and analytics APIs, order-book data and near-real-time WebSocket feeds for third-party builders. Kalshi has begun supplying market data to research buyers, released a beta of Kalshi Pro in July for users managing many markets and resting orders, and reported that institutional trading increased roughly 800% over six months.

The developers describe the tool as a screening layer to help traders spot unusual moves, liquidity and momentum across many contracts.

An unreviewed July working paper that analyzed roughly 23 million Kalshi moneyline trades across NBA, MLB and NHL markets found price calibration was strongest in the middle of a contract’s life and tended to deteriorate in the final 10 minutes before settlement. The paper examined 12,639 filtered parlay trades collected over 15 days and found median pricing near parity for two- to four-leg parlays, with rising overpricing for parlays that had more legs.

A separate unreviewed June working paper on Polymarket’s short Bitcoin contracts identified a spike in Binance spot trading in the final 10 seconds of five-minute contracts and a price reversal after those contracts closed. The study detected a weaker footprint for 15-minute contracts and used Binance as a proxy where contracts were settled against Chainlink; the authors warned against strong causal claims.

The papers note that a displayed probability can be affected by contract wording, settlement method and concentrated trading near the close. The studies suggest treating a single price as one data point and checking contract terms, time to settlement and recent order-flow patterns before interpreting a price as a direct odds estimate.