Prediction Market Probabilities
Prediction-market odds on real events, linked to the stocks each event would move.
- 500+securities mapped
- 300+active events
Sample
2 recordsGeopolitical event to the energy complex (single venue)
Representative shape, modeled on the real record. Six mapped securities with signed exposure and sensitivity 0 to 1.
{
"event_id": "pm-iran-strait-hormuz",
"event_title": "Will Iran close the Strait of Hormuz by March 31?",
"category": "geopolitical",
"source_platform": ["polymarket"],
"current_probability": 0.68,
"probability_7d_change": 0.11,
"volume_total": 8385760,
"observation_timestamp": "2026-03-03T14:30:00Z",
"mapped_securities": [{
"ticker": "CL=F",
"exposure_type": "positive",
"sensitivity_score": 0.93,
"asset_class": "commodity"
}, {
"ticker": "XOM",
"exposure_type": "positive",
"sensitivity_score": 0.82,
"asset_class": "equity"
}, {
"ticker": "SPY",
"exposure_type": "negative",
"sensitivity_score": 0.65,
"asset_class": "etf"
}],
"status": "active"
}Monetary-policy event, cross-platform spread populated
Representative. Tracked on two venues, so cross_platform_spread is set.
{
"event_id": "pm-fed-march-2026-no-change",
"category": "monetary_policy",
"source_platform": ["polymarket", "kalshi"],
"current_probability": 0.97,
"cross_platform_spread": 0.01,
"observation_timestamp": "2026-03-03T14:30:00Z",
"mapped_securities": [{
"ticker": "TLT",
"exposure_type": "negative",
"sensitivity_score": 0.88,
"asset_class": "etf"
}, {
"ticker": "XLF",
"exposure_type": "positive",
"sensitivity_score": 0.72,
"asset_class": "etf"
}],
"status": "active"
}Specifications
| Formats | JSON, CSV, Parquet |
|---|---|
| Delivery | S3, REST API, WebSocket, Email |
| Cadence | Real-time (5-minute resolution) |
| Active events | 300+ |
| Securities mapped | 500+ |
| Platforms | Polymarket, Kalshi, Limitless, Metaculus |
| Resolution | 5-minute continuous |
| History | June 2022-present |
| Access | A derived analytical product. No raw prediction-market data is redistributed. Public-API sourced; no MNPI or PII. Identifiers are also mapped to CUSIP and ISIN. |
What ships
Event Probability Stream
Per-event probability time series with 24h/7d/30d changes, volume, open interest, and cross-platform normalization.
Security Sensitivity Scoring
Event-to-security mapping with sensitivity scores validated against historical price reactions.
Cross-Platform Normalization
Unified schema across Polymarket, Kalshi, Limitless, and Metaculus. Compare identical events across venues.
Fields
13 fields
| Field | Description |
|---|---|
| event_idstring | Canonical event id, unified across venues.Example pm-iran-strait-hormuz |
| event_titlestring | The natural-language question being traded.Example Will Iran close the Strait of Hormuz by March 31? |
| categorystring | Taxonomy: geopolitical, monetary_policy, economic, corporate, election, regulatory, m_and_a, trade_policy.Example geopolitical |
| source_platformstring[] | Venues carrying the event (Polymarket, Kalshi, Limitless, Metaculus).Example ["polymarket","kalshi"] |
| current_probabilityfloat · 0..1 | Latest implied probability.Example 0.68 |
| probability_7d_changefloat | Change in implied probability over 7 days.Example 0.11 |
| volume_totalfloat · USD | Cumulative traded volume.Example 8385760 |
| open_interestfloat · USD · nullable | Open interest where reported by the venue.Example 4200000 |
| observation_timestamptimestamp | Point-in-time stamp at trade-execution time. History is never restated.Example 2026-03-03T14:30:00Z |
| cross_platform_spreadfloat · nullable | Probability divergence when the same event trades on more than one venue.Example 0.06 |
| resolution_datedate | When the event resolves.Example 2026-12-31 |
| statusstring | active or resolved.Example active |
| mapped_securitiesobject[] | The 2 to 8 affected securities: {ticker, name, exposure_type, sensitivity_score, asset_class}.Example [{ticker:"CL=F", exposure_type:"positive", sensitivity_score:0.93}] |
Quality checks
2 checksLead/lag on resolved events
Whether mapped securities subsequently move in the predicted direction after a probability move.
Tested across 500+ historical probability moves greater than 10% in events that have since resolved, spanning a US election cycle, tariff announcements, and Fed decisions; resolved events retained, so no survivorship bias.
Qualitative, reported honestly: probability changes tend to lead the mapped security moves by hours to days. The research declines to publish a single hit-rate because it varies by category, liquidity, and mapping cleanliness.
Per-category sensitivity validation
Whether mapped securities actually move when an event probability changes, by category.
Backtest realized price moves against probability moves with directional correctness, magnitude vs sensitivity score, and lead time, reported per event category.
Methodology-stage. Per-category metrics are computed on request; no fixed published figure is asserted.
Ground truth
| What is correct | The realized security price reaction following a probability move, on events that have resolved. Mappings are validated by asking whether the mapped securities actually moved when the event probability changed. |
|---|---|
| How it is established | For each mapped event-security pair, measure directional correctness on moves greater than 10%, magnitude vs the sensitivity score, and lead time. Cross-platform event matching is verified manually for the top events by volume; probability series are checked for gaps, outliers, and stale prices. |
| Agreement | No inter-rater figure is published; top events are matched under manual QA rather than a rater panel. |
How it is built
- 01
Real-time ingestion
Automated WebSocket and API ingestion from four platforms - one CFTC-regulated exchange, one ICE-backed decentralized exchange, one DeFi market, and one calibrated forecasting aggregator - at 5-minute resolution. No web scraping.
- 02
Cross-platform normalization
Native identifiers (Polymarket condition_id, Kalshi ticker, Limitless market id, Metaculus question_id) resolve to one canonical event, so the same question on multiple venues collapses to one event with multiple probability sources.
- 03
Event categorization
Events are tagged with a standardized taxonomy across geopolitical, monetary policy, regulatory, M&A, election, economic, and corporate categories.
- 04
Event-to-security mapping
Proprietary rules link each event to 2 to 8 public securities with a signed exposure type - trade policy to country and sector ETFs, monetary policy to rates and banks, M&A to target, acquirer, and peers.
- 05
Sensitivity scoring
Each event-security link carries a sensitivity score derived from historical co-movement between probability changes and security price changes.
- 06
Point-in-time stamping
Every observation is stamped at trade-execution time (on-chain settlement or exchange-reported execution). Historical data is immutable, so a strategy at time T sees only observations at or before T.
Request a sample
We send real records in the format you would receive, with the field list and where each record came from. Access for evaluation is limited and covered by a mutual NDA.
ProductPrediction Market Probabilities