The following guest article was submitted by Mikheil Shengelia, Research Analyst at Eagle Alpha, an alternative data aggregation platform providing supporting advisory services for data buyers and vendors. It was written in collaboration with Gerra, a cross-platform prediction market intelligence provider.
Prediction markets have become the fastest-growing source of real-time probability estimates for the political, economic, and regulatory events that move securities. Combined Polymarket and Kalshi volume exceeded $44 billion in 2025; ICE committed up to $2 billion to Polymarket; 73% of institutional market-structure specialists expect tangible value from the data within one to two years.
Prediction Markets as an Event Driven Signal
But a raw probability is hard to act on. A quant covering equities sees “Will Iran close the Strait of Hormuz?” trading at 68¢ and still has to work out which securities it affects, how strongly, and in which direction — then repeat across hundreds of markets on four platforms with completely different schemas. Gerra automates that step: every active event is mapped to the public securities it affects, with directional exposure and a sensitivity score. This note covers Gerra’s data collection and methodology, then focuses on two research questions.
Data Collection and Methodology
Gerra runs their own ingestion rather than redistributing a third-party feed. Market metadata, trade prices, and probability snapshots are pulled programmatically from public exchange APIs and WebSocket feeds at five-minute resolution, normalized into one schema, and backfilled to June 2022. No web scraping.
The layer that makes it tradeable is the event–security map. Every event is categorized into a standard taxonomy and linked to two to eight securities, each with an exposure direction and a sensitivity score (0–1, the predicted price impact per 10% probability change):
- Trade policy → country and sector ETFs, FX pairs; monetary policy → rates instruments, banks, gold.
- Regulatory → directly affected companies and competitors; M&A → target, acquirer, sector peers.
- Point-in-time: every observation is stamped with the source trade-execution time and never restated — no look-ahead in backtests.
- Validation: sensitivity scores are backtested against realized moves across 500+ historical probability shifts of >10% in resolved events, with no survivorship bias
How this differs from what exists. ICE distributes Polymarket-only signals bundled into its broader ecosystem. Gerra is cross-venue (Polymarket, Kalshi, Limitless, Metaculus), normalized, and ships the security-mapping layer on top — built for research and backtesting, not just a real-time terminal widget.
Research Question 1: Do prediction-market probabilities lead the mapped securities around binary catalysts — and over what horizon?
Approach: Take large probability shifts (|Δ| above a threshold), align them to the mapped securities by sensitivity sign, and measure the forward return and the lead time between the probability move and the security move. Three resolved catalysts illustrate the result, spanning an election, a monetary policy decision, and a geopolitical shock.
2024 US PRESIDENTIAL ELECTION → SECTOR ROTATION
Trump win probability on Polymarket rose from ~50% to ~65% in October 2024 on $1.5B of volume — the most liquid political market in history. The mapping identified five securities with directional exposure.
Figure 1: Signal Timeline
When Trump probability crossed 60%, the probability move preceded the equity rotation — XLF +4.2% and ICLN -3.8% over the next five sessions, weeks before Election Day. The mapping called direction on four of five securities.
Figure 2: Schematic of a Delivered Record – One Event-Security Pair. Every event–security pair is one point-in-time record — identical schema across all four venues,
TWO MORE CATALYSTS: A FED DECISION AND A GEOPOLITICAL SHOCK
FED RATE DECISION → BONDS & EQUITIES
The December 2024 rate cut probability jumped from 60% to 85% across Polymarket and Kalshi (combined volume $200M+). The probability shift led the rate-sensitive complex by roughly three sessions.
Figure 3: 5/5 Securities Moved in the Predicted Direction
IRAN GEOPOLITICAL ESCALATION → OIL & DEFENSE
A composite of three live Iran markets (regime change, Strait of Hormuz, US/Israel strikes) mapped to energy, defense, and safe-haven exposure. When the markets surged ~20% in January 2026, the composite signal led the security moves by 6–12 hours.
Figure 4: Mapped Securities Moved in the Predicted Direction
Research Question 2: When the same event prices differently across venues, does that disagreement carry information?
Approach. For events listed on more than one venue, we track the cross-platform spread through time and test whether it predicts how the contract resolves — and whether convergence leads the mapped security. The December 2024 Fed decision is a clean example: the cut probability diverged by three points between Polymarket and Kalshi, then converged inside 48 hours.
Across the dataset, cross-platform divergences greater than 5% resolve within 48 hours more than 80% of the time. That makes the spread both a direct cross-venue convergence signal and a confidence read on the event itself — layered on top of the directional security mapping, not instead of it.
Conclusion
Prediction markets have crossed from curiosity to a real-time, crowd-priced probability layer over exactly the catalysts that move books. The edge for an institutional buyer is not the raw probability — it is a normalized, cross-venue panel with a disciplined event–security map underneath it. This note showed two concrete uses: a lead-lag signal around binary catalysts, and a cross-venue divergence signal that converges on a measurable schedule.
The same panel extends naturally to macro and event-hedging overlays, sector-rotation signals, and tail-risk monitoring — a single view of where prediction markets sit on every name in a book.
In Eagle Alpha’s upcoming webinar on July 9th, we will have an in-depth discussion with AnJie Broad Law Firm on the latest developments in China’s data export regime — covering which data categories remain subject to strict controls, where requirements have eased, and how organizations can assess their compliance obligations when sourcing or utilizing data originating from China. Click here to register.






