In February 2026, Intercontinental Exchange (ICE) launched Polymarket Signals and Sentiment, making it the exclusive institutional distributor of structured data from Polymarket, the world's largest prediction market. The announcement marks a meaningful moment for alternative data: crowd-sourced probability markets are now packaged as a legitimate institutional data feed, sitting alongside securities pricing and fundamental data on ICE's consolidated infrastructure.
This post explains what prediction market data is, how the ICE-Polymarket product works, and how to think about it as an alternative data source.
What prediction markets actually measure
Prediction markets let participants bet real money on the probability of real-world events. On Polymarket, markets cover macroeconomic outcomes (will the Fed cut rates this quarter?), political events, company-specific outcomes, and commodity or currency moves. The market price of a contract reflects the aggregate probability estimate of participants with financial skin in the game.
The argument for prediction markets as information is that they aggregate dispersed knowledge efficiently. Someone who has read deep into Fed commentary, spoken to insiders, or built a rigorous model has an incentive to bet on their view and be rewarded if they are right. That dynamic, the argument goes, produces probabilities that are better calibrated than polls or expert consensus.
The data produced by prediction markets is different from news sentiment or search trends. It is forward-looking probability rather than a reflection of past behavior or current attention. That makes it a complement to behavioral alternative data rather than a substitute.
What ICE launched
ICE's Polymarket Signals and Sentiment product normalizes and structures Polymarket's on-chain data and delivers it through ICE's existing infrastructure:
- ICE Consolidated Feed for near real-time data
- ICE Consolidated History for time-series backtesting
- ICE entity databases to map signals to specific securities and companies
The service uses ICE's entity identification to connect prediction market contracts to publicly traded names and sectors. A market on semiconductor export policy, for example, can be mapped to specific companies in the chip supply chain. A market on consumer confidence can link to retail or housing names.
The product sits within ICE's broader Signals and Sentiment offering, which also includes Reddit Signals and Sentiment and Dow Jones data. The intent is to give institutional investors a structured way to incorporate crowd-sourced and social signal alongside traditional data in a single integration.
What is useful about it
Event-driven and macro research. Prediction markets are most naturally suited to event-driven investing. If you are trading around a Fed decision, a regulatory outcome, or an election, a market that is actively pricing probability provides real-time signal that is harder to get from news sentiment alone. News tells you what happened or what people are talking about; prediction markets tell you what participants currently believe will happen.
Tail risk and scenario work. Even if you are not trading on the exact outcome, knowing the market-implied probability of a scenario helps calibrate risk. A market pricing a 25% probability of a significant policy change is useful context for portfolio construction even if you do not take a view on the direction.
Convergence and divergence. When prediction market probability moves sharply in a direction not yet reflected in equity or options pricing, that divergence can be a signal worth investigating. The same applies when prediction markets and news sentiment point in opposite directions.
Verifiable, timestamped data. Polymarket's on-chain structure means the data is auditable and timestamped. For institutional use, that auditability matters for compliance and backtesting integrity.
What to watch out for
Liquidity varies by market. Prediction market prices are only as good as their liquidity. Thin markets on niche questions can be moved by a small number of participants and may not reflect genuine information aggregation. Liquid markets on high-interest events (Fed meetings, major elections) are more reliable than thin markets on obscure topics.
Not a standalone alpha signal. Prediction market data is directional context, not a trading trigger. Like sentiment data, it is most valuable when combined with other inputs: what does behavioral data (search, traffic) say alongside the market-implied probability? Where are options pricing relative to the prediction market?
Short history for backtesting. Polymarket became liquid enough to be a meaningful data source relatively recently. Backtests on short histories are fragile, and the historical data through ICE will only go back to when Polymarket was active and liquid. Treat early backtests with appropriate skepticism.
Crowding risk. As more institutional participants access the same Polymarket feed through ICE, the edge in the signal may compress. The same happened with early adoption of satellite data, card data, and other alternative feeds. Being early and building proprietary views on top of the raw data matters.
How it fits in an alternative data strategy
Prediction market data addresses a gap that most alternative data does not: explicit probability estimates on specific future events. Other data types tell you about current behavior or historical patterns. Prediction markets tell you what a financially motivated crowd thinks is going to happen.
That makes it most useful in a few specific contexts:
- Pre-event positioning: When you want to understand how much of a macro or policy outcome is already priced
- Scenario calibration: When you want to weight scenarios for portfolio construction or stress testing
- Divergence signals: When prediction markets and other data sources (sentiment, price, options) disagree in a way that suggests mispricing
It is less useful as a general screen for stock ideas or as a substitute for behavioral data. Search trends, news sentiment, and web traffic capture what is happening now in consumer or market behavior. Prediction markets capture what people expect to happen. Both are valuable, and they answer different questions.
Platforms that cover multiple data types, including search, sentiment, and social, alongside event-driven feeds give investors a more complete picture. For search, sentiment, and behavioral signals alongside event-driven research, see Paradox Intelligence and Research.
Bottom line
The ICE-Polymarket launch is the clearest signal yet that prediction market data is moving from a niche curiosity to a mainstream institutional input. The product provides structured, verifiable, entity-mapped probability data from the world's most liquid prediction market. Its value lies in event-driven and macro research, scenario probability, and divergence signals rather than in general stock screening. Used as one layer in a multi-signal process, it adds something that no other data type offers.
For more context on the alternative data landscape in 2026, see Best Alternative Data Platforms 2026.
This post is for institutional investors and research professionals. It is not investment advice.
