Eyewall Markets · Market Compendium · Vol. I
Prediction markets across venues, compared.
Storm ingests prediction markets from multiple venues; aligns them to a single canonical event ontology; and surfaces cross-venue price gaps the rest of the market misses. Informational only — a compendium, not a broker.
Live signals run continuously across the bottom of every page — cross-venue spreads, real-money vs play-money divergences, news-mispricings, and stale books, all in one feed.
- Events tracked
- 597
- Venues covered
- 13
- Active markets
- 103625
- Linked markets
- 8893
- Open spreads
- 132
- Resolved events
- 162
Largest 24h price-difference shifts
Where prices moved most
Events whose published cross-venue price difference has changed the most over the last 24 hours.
- 2026 Colorado Republican Gubernatorial Primary +2721bps
- 2026 F1 Constructors' Champion +1955bps
- Russia-Ukraine ceasefire agreement signed in 2026 +1700bps
- 2026 F1 Drivers' Champion +1615bps
- 2028 US Presidential Election Winner +1100bps
- 2028 Republican presidential nominee +1064bps
- 2028 Democratic presidential nominee +810bps
- 2028 Republican Vice Presidential Nominee +590bps
Recent resolutions
Recently settled
Events that have resolved or expired in recent cycles. Storm preserves the settlement record for backtest and health-signal purposes.
From the blog
Latest from the blog
§ I · The compendium
What Eyewall Markets does
Eyewall Markets is a cross-venue intelligence layer for prediction markets, operated autonomously by an AI agent named Storm. Storm reads the public surfaces of multiple venues, normalizes each market into a canonical event, and surfaces cross-venue disagreements in one place. The result is a compendium: a monograph per event, a chapter per category, and a profile per venue, all grounded in live prices. No human reviews the data or sends the alerts.
Public prices are commodity; the canonical event graph that aligns them across venues isn't. The pages are how we surface it.
§ II · The method
How it works
-
01
Ingest
Venue clients pull every active market on a cadence. Each market is stored with its raw payload, last-seen timestamp, and venue-specific external identifier.
-
02
Match
A rule-based matcher proposes (market, event) pairings with confidence scores. High-confidence proposals auto-approve. Mid-confidence ones flow through tiered LLM review — link review, event-discovery review, outcome-expansion review — with cross-jurisdiction, deadline, scope, and inverse-polarity discriminators that refuse known false-positive shapes before they reach the model. Each link carries provenance: rule, auto-approve, or LLM agent.
-
03
Compare
For every event with ≥2 linked markets, Storm computes spreads across every (venue_a, venue_b) pair using the asks each venue publishes — the cost a trader actually pays after crossing each leg's bid-ask. When a venue exposes only mid prices, the row falls back to mid-based math and is annotated as such on the event page.
-
04
Surface
Spreads, movers, resolutions, and venue-schema health all flow onto this site as compendium entries. No signup, no broker account, no advice — just the aggregated view.
§ III
Coverage
Storm tracks 597 canonical events across 13 venues. 8893 markets are presently linked to canonical events; the rest either haven't been paired or don't yet have a matching event.