Pillar 2: Infrastructure

Prediction market infrastructure spans a spectrum from fully onchain to fully offchain systems, with hybrid models increasingly becoming the standard. Fully onchain systems maximize transparency and decentralization but face scalability constraints, while offchain systems prioritize efficiency at the cost of trust assumptions.

Hybrid architectures balance these trade-offs by combining offchain order matching with onchain settlement. In this model, user orders are processed offchain and subsequently executed via smart contracts, enabling both performance and verifiability.

Pillar 3: Resolution

Resolution mechanisms determine how outcomes are verified and settled, making them critical to overall system trust. Even highly efficient pricing systems lose credibility if outcomes are resolved inaccurately or inconsistently.

Resolution operates across two layers: governance mechanisms that define truth criteria, and oracle systems that deliver verified data onchain. Centralized governance approaches such as human judgement prioritize speed and simplicity, while decentralized jury models and hybrid systems enhance resistance to manipulation at the cost of complexity.

Oracle infrastructure similarly varies in design. Centralized oracles provide fast data delivery but introduce single points of failure, whereas decentralized networks like Chainlink’s Decentralized Oracle Networks (DONs) aggregate inputs from multiple independent node operators, each with staked economic incentives for accuracy, and delivers cryptographically signed, timestamped price reports that smart contracts can verify onchain without trusting any single source.

Its October 2025 integration with Polymarket — combining Chainlink Data Streams with Chainlink Runtime Environment (CRE) — enabled near-instant, automated resolution of asset pricing markets on Polygon mainnet, eliminating human delay and dispute risk entirely for objective market types. For markets tied to deterministic asset prices, economic data releases, and quantifiable benchmarks, Chainlink's infrastructure makes trustless, automated settlement possible at scale.

For more subjective markets, Chainlink’s emerging AI oracles can help automate resolution by sourcing and verifying facts across diverse inputs, then using consensus mechanisms to improve transparency and reliability for long-tail, high-context questions.

A hybrid derivative

Given their design, prediction markets are simultaneously similar to options, futures, and bonds. Prediction market contracts often pay $1 (face value) if an event occurs, resembling binary options with a premium paid for contingent payoff and futures through pre-collateralized, binding settlement at a fixed maturity (resolution date). They also borrow bond-like traits with defined payouts, but event-based instead of interest payments.

Due to such hybrid nature, prediction markets defy clean regulatory classification, as evidenced by CFTC struggles. The agency classifies them as unregistered "event contracts" or binary options/swaps under the Commodity Exchange Act, yet their event-driven oracles (e.g., elections) fall outside commodity norms, sparking gaming-vs-derivatives jurisdictional fights that delayed approvals for years, with only limited contracts allowed on platforms like Kalshi. Polymarket's 2022 $1.4 million CFTC fine and U.S. user ban exemplify this, as its crypto peer-to-pool model blurred swaps and gambling lines without DCM registration.

In order to continue operations in the U.S., Polymarket acquired CFTC-licensed exchange QCEX in July 2025. As of March 2026, cumulative notional volume in Polymarket U.S. reached $1.25 billion, with a weekly high of $248.4 million in the last week of March.

Exchange-like fee structure

Kalshi operates under an exchange-based model, charging transaction fees on each trade by taking a percentage of expected returns from both buyers and sellers. Unlike traditional sportsbooks, the platform does not take directional exposure against users, but instead facilitates trade matching between participants. As a result, its revenue is directly linked to trading activity, similar to how traditional exchanges monetize volume.

Users driving the market

Prediction market accuracy is driven by the same incentive structure underlying financial markets: profit-seeking behavior. Participants who perceive the true probability to be higher than the market price are incentivized to buy, pushing prices upward, while those who assess it as lower are incentivized to sell, driving prices downward. Unlike polling systems, where responses carry no financial consequence, positions in prediction markets are backed by capital at risk. This structure rewards accurate assessments and penalizes mispricing, reinforcing the integrity of price signals.

This incentive alignment has contributed to sustained user growth. As of March 23, total unique users across prediction markets exceeded 2.8 million, with Polymarket accounting for 79.6% of the total. Kalshi recorded over 74,000 cumulative onchain users. Jupiter, a Solana-based application integrating prediction markets, has also experienced rapid adoption, with its user base surpassing 15,000 within six months and continuing to expand steadily.

In traditional finance, risk management models are primarily constructed around volatility to estimate asset price movements and inform hedging strategies. Early frameworks such as Black–Scholes assume constant volatility and continuous price evolution, while more advanced models incorporate discontinuities and sudden market shocks. Despite these improvements, the modeling of discrete, binary event risk remains constrained. Existing approaches typically rely on historical data, limiting their forward-looking capability, or infer event probabilities indirectly through market proxies, introducing noise and inefficiency.

Prediction markets address this limitation by directly pricing event probabilities in real time. The resulting outputs provide clear, continuously updated probability estimates that can be integrated into existing risk models without additional transformation.

Accurate discrete event probabilities

Empirical evidence supports the accuracy of these signals. Research conducted by Federal Reserve economists finds that Kalshi’s forecasts for the federal funds rate and CPI deliver statistically significant improvements over fed funds futures and professional forecasters, while also providing continuously updated probability distributions rather than static point estimates.

Additional analysis comparing Polymarket data with traditional polling during the 2024 U.S. election highlights similar advantages. Polymarket not only produced more accurate outcome predictions, but also responded more rapidly to major events, including presidential debates, shifts in candidate participation, and other key developments throughout the election cycle. At the state level, its forecasts for most major swing states also outperformed polling data. By mid-October 2024, as open interest in election markets expanded, Polymarket probabilities consistently indicated a Trump victory, while polling data continued to favor Harris.

In fact, Polymarket demonstrates consistently high predictive accuracy, reaching an average of 90.8% four hours prior to market resolution and 86.2% one month in advance. Evaluation through Brier scores further reinforces this performance, with lower scores indicating greater accuracy. Between June 2024 and March 2026, the cumulative Brier score — measured in half-month intervals — exceeded 0.1 only once, in mid-October 2024, before trending downward through 2025 and 2026 to an overall score of 0.0843. This decline reflects a steady improvement in predictive performance over time. The lowest recorded half-month score occurred on the fortnight of January 1, 2025, at 0.0048, indicating exceptionally high forecast precision during that period.

Institutional and enterprise use cases

Institutional engagement with prediction markets has progressed beyond passive observation into active integration. By early 2026, leading hedge funds and quantitative trading firms, including Oldenburg Capital Partners, had incorporated prediction market data directly into risk modeling frameworks. Similarly, Saba Capital has reportedly utilized recession-linked contracts on Polymarket to hedge credit exposures that may lag behind evolving macroeconomic conditions.

Adoption is also expanding into adjacent sectors such as insurance. In February, sports insurance broker Game Point Capital announced a partnership with Kalshi to hedge NBA performance-based bonuses at nearly half the costs of traditional reinsurance markets. Initial transactions executed prior to the announcement illustrate this efficiency, with pricing significantly below comparable over-the-counter rates (6% versus 12-13%). In parallel, Kalshi established a strategic partnership with Tradeweb to embed probabilistic, forward-looking risk signals into trading workflows used by over 3,000 institutional clients, further integrating prediction markets into core financial infrastructure.

Survey data indicates growing interest within the proprietary trading community. A late-2025 study by Acuiti conducted through its Proprietary Trading Expert Network, found that 10% of firms are already active in prediction markets, with an additional 35% evaluating participation. In the United States, adoption is more pronounced, with approximately three-quarters of firms either trading or considering entry compared to 37% in Europe. Forward-looking sentiment is similarly strong: only 6% of respondents indicated that prediction markets would not play a role in the institutional proprietary trading landscape over the next three to five years.