Once your Resolution Spec is tight, the next fork is how prices form. A continuous limit-order book and an automated market maker can both list the same event, but they create different first-week experiences, different manipulation surfaces, and different compliance narratives. Wrong engine choice does not create bad forecasts by itself—it creates fake liquidity, untradeable size, and traders who learn to distrust your mid on day one.
Mechanism is a stage-of-life decision, not a logo preference.
The AMM versus order book comparison chapter in the market-structure module gave trader-facing intuition; here the question is operational: who you must pay, what you must monitor, and what honesty your UI owes on day one.
How traders experience each engine
On a book, the trader sees bid, ask, spread, and depth at a glance—friction is visible. On an AMM, the trader often sees one mid until size is entered; friction hides in the curve. Builder UX should make effective price at the trader’s typical clip as prominent as the mid tile. That single design choice prevents a class of “rigged app” reviews on launch week.
Order book markets in plain terms
In a central limit-order book, buyers and sellers post bids and offers; the exchange matches them by price and time priority. Depth is visible at the touch; large clips can walk the book with predictable slippage if makers are present. Strengths are credible price discovery at size, familiar UX for pros, and a story regulators understand when surveillance and market-maker obligations exist. Weaknesses are brutal on launch day: an empty ladder looks like a 40¢/60¢ joke, and you must pay for makers, matching infrastructure, and integrity monitoring.
This path fits marquee events with expected flow, jurisdictions that expect DCM-style market integrity, and teams that can run twenty-four-seven market operations.
AMM markets in plain terms
An automated market maker quotes a price from a formula and pool reserves (or an LMSR liquidity parameter). Strengths are always-on mids, permissionless listing on crypto stacks, and long-tail coverage without begging a desk for quotes. Weaknesses are capital hunger, curve manipulation when total value locked is tiny, and mids that lie about cost for anything beyond a small clip. Retail sees 50¢; a five-thousand-dollar buy might pay 54¢ effective—your UI must show that honestly.
This path fits cold start, wallet-native UX, and catalogs where many niche events matter more than one perfect touch.
Hybrids are a migration story, not a cheat code
Mature venues often run hybrid stacks: an order book for headline size, an AMM backstop for odd lots, sometimes RFQ for blocks. The product job is labeling which leg filled and which mid the tile displays. Traders should not discover after the fact that their “market price” was a thin pool print while the real depth lived on a book they never saw.
Plan the migration before launch hype anchors everyone to a misleading formula mid. A common arc is seed an AMM, recruit makers when seven-day notional crosses a threshold you define, then route large clips to the book while the pool absorbs tail flow.
Matching mechanism to contract shape
Simple binaries work on either engine. Categorical markets with five named nominees need either an exchange-style book with cross-outcome market makers or an AMM design that enforces mutual exclusivity—five unrelated puddles whose prices sum to 130% teach traders you do not understand the product. Scalar and conditional markets add inventory and dispute gamma; launching a conditional child on a solo pool with two thousand dollars of TVL is how you manufacture orphan liquidity and dual disputes.
Complexity in contract shape multiplies mechanism risk. Do not ship a bundle on infrastructure meant for a single binary.
Stage-based choices
On day one of an unknown niche approval market, an empty CLOB is honest but ugly; a subsidized LMSR prints a mid immediately at the cost of treasury burn. At viral headline scale, thin AMM depth is a credibility trap while a staffed CLOB with makers can absorb six-figure clips near the touch. Between those extremes lives the hybrid transition: pool for coverage, book for size, emissions that sunset on a schedule you publish.
Think in weeks, not ideology: what touch do you need before journalists quote you?
Numeric comparison at launch week
Imagine a niche regulatory approval binary with clean rules but unknown volume. A CLOB-only launch might show empty or eight-cent-wide markets until makers arrive. A low-liquidity-parameter LMSR might print 50¢ with roughly four cents effective slippage on a five-thousand-dollar buy if you subsidize the curve. A documented hybrid costs more engineering but can land near three cents if you seed the pool and post a thin book.
For a national election binary with millions in notional interest, a thin AMM is a credibility trap: a hundred-thousand-dollar order might move ten percent. A CLOB with paid makers is the serious lane; AMM, if present, is for retail clips—not for the headline trade.
Dispute halts are not mechanism-agnostic
When an outcome is asserted or challenged, open orders must cancel and swaps must pause. Hybrid venues need a runbook for which leg halts first and how liquidity providers are informed. Settlement of winning tokens can be standard; trading halt timing is not. LP exposure frozen through a dispute is inventory risk plus legal risk—price it into fees and subsidies.
Regulated US vs crypto-native framing
US-facing retail dollars often push teams toward supervised CLOB narratives: surveillance exports, maker accountability, tick sizes. Crypto-native global stacks favor AMM-first listing speed and wallet flows, accepting gas UX and token-policy overhead. The launch-versus-build chapter intersects this choice; here the point is to pick the engine that matches who may trade and how they fund, not the brand you envy on social media.
Failure modes builders recognize late
Fake liquidity is an eight-hundred-dollar pool with a tight mid—hide or widen until TVL crosses a threshold you publish. Ghost books quote one lot at the touch. Hybrid miswiring creates arb between your own legs. Retail rage follows curve cliffs you did not simulate. Regulatory scrutiny arrives when you claim exchange-grade integrity on unsupervised matching.
Disclose engine and slippage honestly. Efficiency is conditional on depth and rules, as the foundations module stressed.
Fees, ticks, and surveillance (CLOB specifics)
Tick size and minimum lot matter for manipulation economics. A one-lot quote at the touch is not depth—it is a billboard. Surveillance for spoofing, layering, and wash prints is part of the product on regulated paths, not an afterthought. If you cannot staff surveillance, do not market exchange-grade integrity.
Pool parameters (AMM specifics)
Liquidity parameter b in LMSR-style designs sets how much a marginal trade moves price. Lower b means tighter mids and faster manipulation on small bankrolls. Higher b means flatter curves and more capital to achieve tight touch. Publish the parameter or pool weights; let traders simulate clips before they click buy.
Choosing after you have real volume data
Revisit mechanism quarterly. A market that never cleared five thousand dollars notional might belong in a pool forever; a market that clears millions on debate night might need a book leg within weeks. Migration is a communications problem as much as an engineering problem—traders anchored to a pool mid need notice before you route them elsewhere.
Surveillance and halts (both engines)
News halts should be defined in the rulebook: cancel resting orders on CLOB, pause swaps on AMM, show a single status banner. Ad hoc pauses during disputes without comms train traders that rules are optional.
Key ideas
Pick mechanism for volume stage, contract shape, and jurisdiction. CLOB buys credibility at size; AMM buys coverage at birth; hybrid buys honest scale transitions. Plan dispute halts per leg. Never market a mid your simulator cannot defend at real clip sizes.
What comes next
You chose how traders negotiate odds. Next you choose who may certify truth and what bonds make lying expensive—oracle and dispute are one product, not two vendors picked from a menu.
Run a table-top: a five-thousand-dollar market order on your launch config—write down effective price, fees, and halt behavior. That number is what honest marketing may cite.
Next: Selecting an Oracle and Dispute Resolution Framework