Chapter 5: Key Takeaways

Platform Landscape

  1. The modern prediction market ecosystem spans four quadrants: regulated real-money (Kalshi), crypto-native real-money (Polymarket), play-money (Manifold Markets), and forecasting platforms (Metaculus). Each quadrant serves different users and use cases.

  2. Polymarket is the volume leader, processing hundreds of millions to over a billion dollars in monthly trading volume, powered by the Conditional Token Framework on Polygon with a hybrid on-chain/off-chain CLOB. It is the platform of choice for high-liquidity trading but is officially unavailable to U.S. residents.

  3. Kalshi is the regulatory pioneer, holding CFTC Designated Contract Market status — the same designation as the CME. It provides the strongest consumer protections and legal clarity for U.S.-based traders, though at the cost of slower market type expansion and geography restrictions.

  4. Metaculus is not a market but a forecasting platform that supports continuous probability distributions, conditional questions, and rigorous scoring. It excels for long-term questions and produces remarkably well-calibrated aggregate forecasts despite using only points as incentives.

  5. Manifold Markets democratizes prediction markets by letting anyone create a market on any topic. Its play-money model (Mana) eliminates regulatory barriers and produces surprising accuracy through reputation incentives, though market quality varies widely.

  6. PredictIt pioneered U.S. real-money prediction markets under a CFTC no-action letter, but its constraints (850-trader limit, $850 position cap, high fees) limited its effectiveness. Its legacy lives on in the regulatory precedents it established.

Technical Architecture

  1. The Conditional Token Framework creates complementary token pairs (Yes + No = $1) that can be freely traded. This elegant design enforces the probability constraint and enables non-custodial markets.

  2. CLOBs and AMMs each have trade-offs: CLOBs (Polymarket, Kalshi) offer tighter spreads and better price discovery but require market makers. AMMs (Manifold) provide guaranteed liquidity but with wider spreads and different price impact characteristics.

  3. Market resolution mechanisms vary significantly: from centralized teams (Kalshi) to oracle protocols (Polymarket/UMA) to creator-based resolution (Manifold). Each involves trade-offs between speed, reliability, and decentralization.

Practical Considerations

  1. Fee structures dramatically affect profitability: PredictIt's combined 10% profit fee and 5% withdrawal fee could consume 15% or more of gross profits. Polymarket and Kalshi have much lower fees, making them more suitable for active trading.

  2. Cross-platform price discrepancies are common and persistent, ranging from 5-10 percentage points for the same event. These arise from geographic segmentation, different user bases, fee structures, and friction in moving capital between platforms.

  3. True cross-platform arbitrage is usually impractical due to regulatory barriers (U.S. vs. non-U.S.), different currencies (USD vs. USDC), capital lockup, and potential differences in resolution criteria.

  4. All major platforms offer public APIs with varying levels of access. Manifold is the most open (fully open-source), while Kalshi requires authentication for most endpoints. API rate limits range from approximately 30 to 600 requests per minute.

Choosing a Platform

  1. Match platform to purpose: Use Polymarket or Kalshi for trading profit, Metaculus for forecasting practice, Manifold for learning and experimentation, and any platform with good APIs for research and data analysis.

  2. Regulatory risk remains real: The CFTC's treatment of prediction markets continues to evolve. Kalshi provides the most regulatory safety for U.S. users. Non-U.S. users face less regulatory friction but also fewer protections.

  3. Combining data from multiple platforms typically produces better forecasts than relying on any single source. Each platform's user base brings different information and biases, and cross-platform consensus is a strong signal.

Chapter 5 of "Learning Prediction Markets — From Concepts to Strategies"