Automated market makers (AMMs) quote prices from math and pool state, not from a ladder of human limit orders. Crypto prediction markets popularized them; academic designs such as LMSR predated DeFi.
This chapter explains the idea before the formulas in later chapters. Contrast a regulated touch built from bids and asks with a pool that prints “63% YES” until arbitrageurs and limit orders pull it toward external consensus.
What problem AMMs solve
Order books need someone to post bids and asks. For new or thin event contracts, nobody shows up—prices do not exist. AMMs promise an always-on price (if the pool has reserves), permissionless liquidity from users who deposit capital, and simple UX—“swap YES for stablecoin” like a vending machine.
You trade against a pool, not against another person’s resting order, though arbitrageurs bridge pools to books when both exist. The formula can still return a price on an empty CLOB day; that does not mean the price is an informative forecast unless size backs the print.
Historically, academic LMSR markets and corporate forecasting pilots used AMMs before crypto made them visible to retail. DeFi generalized the pattern to wallet-based swaps; prediction markets adopted it because listing thousands of niche questions on separate CLOBs was economically impossible without automated quotes.
Pool-based intuition
Imagine a pool holding YES and NO shares (or YES and cash). A pricing rule says more buyers of YES push the YES price up; more buyers of NO push YES down. The rule is coded in a curve (LMSR, constant-product, and others). Slippage is automatic: large trades move the curve.
A balanced pool might open near 50%. A $500 YES buy might lift implied probability to 52%; a $3,000 YES buy might reach 58% with noticeable slippage. If an external book shows 55%, arbitrage pressure can pull the pool back toward that anchor—when legs are cheap enough and rules match.
AMM vs book in one breath
On an order book, price comes from competing quotes and spread from maker competition; depth comes from posted limits. On an AMM, price comes from formula plus inventory, spread and slippage from curve shape, and depth from pool size and the liquidity parameter b discussed later. Manipulation cost is often lower when the pool is small—always a price, not always an honest one.
Who provides liquidity?
Liquidity providers deposit capital into the pool. They earn fees on volume but take inventory risk—if YES rallies, they may hold losing NO exposure. Traders pay fees and move prices when they swap. Arbitrageurs align AMM price with external books; without them, a pool percentage can drift from a regulated mid for minutes.
Constant-product and cousins
Not every pool is LMSR. Constant-product curves (familiar from token swaps) can price binary outcomes with extra wiring so probabilities sum to one. The economic story is the same: inventory in the pool determines price, and large trades move the curve. Later chapters focus on LMSR because it is the cleanest teaching model for probability events; live venues may blend or modify formulas.
Where you see AMMs
Early on-chain politics markets, play-money and research platforms using LMSR, and hybrid venues that use a pool baseline plus limit orders on top. Regulated U.S. CLOB venues are not pure AMM—but understanding pools explains why crypto headlines quote a single percent while regulated apps quote bid and ask.
Settlement is separate from the curve
The AMM answers “what price for the next trade?” The oracle or exchange rulebook answers “who gets paid at the end?” Confusing the two leads to trades that made sense intrade but were worthless at resolution because the event definition differed from your model. Pool math can be perfect while the contract is still untradeable for forecasters if resolution is vague—carry the opening module’s resolution discipline forward.
Benefits and costs
Benefits include continuous quotes on long-tail markets, fast listing of new questions, composability with wallets and stablecoins, and no queue-priority games for retail. Costs include slippage on big trades, subsidy and capital lock for deep liquidity, oracle settlement separate from the pricing engine, and manipulation that is cheaper when the pool is small.
Retail sees “63% YES.” Under the hood the AMM moved along a curve; the next $10,000 buy might print 67%. Always ask pool size? A percentage without depth is decorative.
Price ≈ probability on AMMs
A displayed $0.63 YES still means roughly 63% if the curve is liquid enough that your size does not move it materially. Unlike a CLOB mid between 61¢ bid and 65¢ ask, the AMM often shows one number that embeds tiny-trade slippage—but not your $20,000 block. Your average fill is the probability relevant to your P&L, not the pre-trade banner.
If pool YES is 67% and a regulated book’s YES ask is 60%, bots sell expensive YES or buy cheap elsewhere until frictions stop them. Persistent gaps are where later modules on arbitrage begin: capital controls, wallet friction, and mismatched event text all show up as “free money” that is not free. Pumping a thin pool from 50% toward 80% may cost hundreds of dollars of slippage the pumper owns until someone takes the other side; on a deep book the same headline move costs more—AMMs are double-edged.
Retail UX versus institutional reality
The swap button hides the curve, which is good for adoption and dangerous for size. Platforms that care about serious forecasting add trade preview, impact charts, and pool TVL in the same screen as the percent. When those are missing, assume you are the product demo, not the informed counterparty.
Fees and composability
Pool trades often pay a protocol fee to LPs plus gas on-chain. Fee yield is what attracts LPs to absorb inventory risk; without volume, the pool is a parked subsidy. Composability with wallets and stablecoins lowers onboarding friction but does not replace disclosure: you still need TVL, volume, and resolution text in the same mental frame as the percent.
Common misconceptions
The AMM price is not “true odds” in isolation—it is the marginal price at current inventory, and size moves it. LPs are makers with different risk. Pools can be arbed against books daily when rules align. Small liquidity parameter does not mean “tight spread”; it often means easy to move.
Many scaled platforms run pool for discovery plus book for size—the hybrid chapter picks up that architecture.
Inventory risk in plain language
When you buy YES from the pool, the pool often ends up shorter YES and longer NO (or cash) exposure. Someone must hold the other side until resolution—that is LP inventory risk. Fees are partly payment for warehousing that risk. If you would not run a book stall without compensation, do not expect LPs to absorb your information for free.
When not to use an AMM
If a regulated book exists on the same event with tight spread and you are trading more than a few hundred dollars, start on the book unless fees or access block you. If no book exists and TVL is tiny, consider not trading at all rather than trading the curve. AMMs expand access; they do not remove the need for judgment.
When you evaluate a pool market, ask three questions in prose, not checklists: Is there enough capital that my trade is a small perturbation? Is someone arbitraging this print to a regulated or sports-book anchor? Is resolution identical if I compare across venues? Weak answers mean treat the percent as local color, not national forecast.
Relationship to Module 01
Module one defined events, contracts, resolution, and price-as-probability. AMMs do not change that vocabulary—they change who posts the first price and how size moves it. Keep resolution discipline; add pool discipline. The next chapters derive LMSR and walk the click path; this chapter only establishes that pools are a different liquidity technology, not a different definition of prediction market.
Summary sentence
AMMs sell availability of a price; they do not by themselves sell quality of a forecast—that still requires capital, flow, and clear resolution.
Key ideas
AMMs trade against a pool and formula, not a human quote stack. They solve cold start at the cost of curve slippage. LPs supply capital and absorb risk; fees compensate them partially. Read pool markets and book markets as coexisting in one ecosystem—compare execution, not screenshots.
Pools trade certainty of quote for uncertainty of depth—that is the deal you accept when you skip the ladder.
Next: the classic academic engine—LMSR (Logarithmic Market Scoring Rule).