Part V: Market Design & Mechanism Engineering
Chapters 28--33
Up to this point, the book has treated prediction markets from the perspective of a participant -- someone who trades in markets, analyzes their data, and builds models to forecast their outcomes. Part V shifts the vantage point entirely. Here, you become the architect. You are no longer asking "How do I profit from this market?" but rather "How do I design a market that produces accurate forecasts, attracts participants, and sustains itself over time?"
This shift in perspective is more than an intellectual exercise. The prediction market ecosystem is still young, and the platforms that exist today represent only a narrow slice of what is possible. Many of the most impactful applications -- decision markets inside organizations, combinatorial markets for complex supply chains, public policy forecasting platforms -- remain underbuilt or unbuilt. If the ideas in this book excite you, there is a reasonable chance you will someday find yourself designing a prediction market, whether as a product manager at a tech company, a researcher at a policy institute, or a founder building the next major platform. This part equips you for that role.
Chapter 28 begins with question design, the single most consequential decision a market creator makes. A poorly worded question -- one that is ambiguous, unfalsifiable, or misaligned with the information need it is meant to serve -- will produce a market that is noisy, contentious, and uninformative regardless of how sophisticated its mechanism is. We develop a systematic framework for crafting resolution criteria, choosing time horizons, and stress-testing questions against edge cases. You will analyze real examples of both well-designed and catastrophically ambiguous markets.
Chapter 29 addresses liquidity provision, the lifeblood of any functioning market. A market without liquidity is a market without information. We examine the economic incentives of market making, the role of subsidized liquidity in bootstrapping new markets, and the practical mechanics of providing liquidity on both centralized and decentralized platforms. You will learn to calculate the cost of subsidizing a market and to evaluate whether that cost is justified by the informational value produced.
Chapter 30 ventures into combinatorial prediction markets, which allow participants to trade on combinations of outcomes across multiple questions. These markets are extraordinarily powerful in theory -- they can capture conditional probabilities and complex dependencies that simple markets cannot -- but they pose formidable computational and design challenges. We survey the key mechanisms (combinatorial LMSR, market making with constraints) and discuss which applications justify the added complexity.
Chapter 31 explores one of the most provocative applications of prediction markets: decision markets and futarchy. The core idea, due to Robin Hanson, is to use prediction markets not merely to forecast events but to guide decisions. We examine the theoretical foundations, the practical attempts (including corporate decision markets and blockchain-based governance experiments), and the legitimate objections. This chapter asks you to think carefully about when collective intelligence should and should not be delegated to a market mechanism.
Chapter 32 is the most hands-on chapter in this part: building a prediction market platform from scratch. We walk through the full stack -- data model, matching engine, market maker, user interface, and resolution system -- using modern web technologies. This is not a toy exercise. By the end of the chapter, you will have a working prototype capable of hosting real markets, and more importantly, you will understand the engineering trade-offs that every platform team faces.
Chapter 33 takes that prototype and confronts it with the realities of scale. What happens when your market has ten thousand concurrent users instead of ten? How do you handle abuse, manipulation, and disputes? What are the operational, legal, and community-management challenges of running a live platform? We draw on case studies from existing platforms to map the terrain between "working prototype" and "sustainable product."
By the end of Part V, you will understand prediction markets not just as tools to use but as systems to build. You will have the design vocabulary, the engineering skills, and the practical awareness to create markets that are well-specified, adequately liquid, and structurally sound -- markets that deserve the trust their participants place in them.