How Are CFMMs Equivalent to Prediction Markets?

TL;DR

Every well-behaved CFMM with a concave, increasing potential function can be converted into an equivalent cost-function prediction market with one outcome per asset, and the conversion also works in reverse. This equivalence connects trade facilitation with information elicitation, showing that the same desirable market-making properties can support both asset exchange and the revelation of traders’ beliefs.

Transcript

so welcome everyone it's been a while but uh a16z crypto research seminars back for today uh very happy to introduce Beau Wagner professor at University of Colorado um telling us about his new work on off the press um about amm's and prediction markets and the connections between them so about all yours great thanks great so this is... Read More

Key Insights

• A CFMM is an automated exchange that facilitates trades among a collection of assets through a single market maker, with traders interacting sequentially with that market maker rather than trading directly with one another.
• A trade is represented as a vector describing the market maker’s net position change, where a positive component means the market maker receives an asset and a negative component means the market maker transfers an asset to the trader.
• Current reserves are determined by adding all completed trade vectors to the market maker’s initial reserves, which are assumed to contain positive quantities of every asset in the model.
• A constant-function market maker accepts a proposed trade exactly when its potential function has the same value for the reserves before and after that trade, keeping the market on a designated level set.
• The constant-product market maker is a primary CFMM example in which the product of asset quantities remains constant as accepted trades move the reserve position along the corresponding valid-reserve curve.
• Four market-making axioms characterize CFMMs whose potential functions are concave and increasing, providing an axiomatic explanation for why this familiar class of mechanisms represents desirable automated market makers.
• Every qualifying CFMM on n assets is equivalent to a cost-function prediction market for an event with n outcomes, and the presented construction can translate mechanisms in either direction.
• Desirable trade-facilitation axioms correspond to desirable information-elicitation axioms, so a market’s ability to support beneficial exchange is technically linked to its ability to reveal participants’ beliefs.

Questions & Answers

Q: What is a constant-function market maker?

A constant-function market maker is an automated exchange that offers trades among a set of assets through one market maker. It maintains a potential function over its reserves and accepts a proposed trade exactly when the function’s value is unchanged after the reserves are updated. Valid trades therefore move the reserve position between points on the same level set.

Q: How does a CFMM decide whether to accept a trade?

A CFMM compares its potential function at the current reserves with the function’s value at the reserves that would result from the proposed trade. The trade is acceptable if and only if those two values are equal. This rule keeps the market maker’s reserve state on the same valid-reserve curve throughout the sequence of accepted trades.

Q: How are trades and reserves represented in the CFMM model?

A trade is represented by a vector with one component for each asset, recording the market maker’s net position change. A positive component means the market maker gains that asset, while a negative component means it transfers the asset to a trader. Current reserves equal the initial reserves plus the sum of all trade vectors completed so far.

Q: What axioms characterize desirable constant-function market makers?

The research states that an automated market maker satisfies a particular collection of four axioms if and only if it is a CFMM represented by a concave, increasing potential function. The transcript does not provide the complete formal list in the supplied portion, but it explains that the axioms narrow a very general, potentially history-dependent pricing rule to the familiar CFMM class.

Q: What is the relationship between CFMMs and prediction markets?

Every CFMM in the characterized class operating on n assets can be converted into an equivalent cost-function prediction market for an event with n outcomes. The relationship is bidirectional, since the construction also converts a suitable prediction market into a CFMM. This establishes a technical equivalence between two market designs developed for apparently different purposes.

Q: Why does the CFMM and prediction-market equivalence matter?

The equivalence shows that desirable properties for facilitating trades correspond to desirable properties for eliciting information. CFMMs are commonly viewed as mechanisms for exchanging assets, while prediction markets are designed to reveal beliefs about uncertain outcomes. The result demonstrates that the same underlying market-making structure can serve both goals and allows technical tools from either field to interoperate.

Q: What is the constant-product market maker example?

The constant-product market maker is a CFMM whose accepted trades preserve the product of the quantities held in reserve. With two assets, the reserve state lies on a curve containing every pair of quantities with the designated constant product. Trading changes how much of each asset the market maker holds while keeping the reserve position on that curve.

Q: What limitations does the presented market-maker model impose?

The discussion deliberately studies a small, closed system containing one market maker and sequentially arriving traders. It excludes liquidity providers and interactions with other markets that might display different prices. Assets are assumed to have non-negative value, and the core characterization focuses on a vanilla model, while practical modifications such as transaction fees can weaken path-independence properties.

Summary & Key Takeaways

• A constant-function market maker facilitates sequential trades among assets through a single automated intermediary. Its state is represented by current reserves, calculated from initial reserves plus completed trades. A trade is accepted when the potential function has the same value before and after the proposed change in reserves.

• The research begins with a very general automated market-maker model, then applies four axioms to characterize CFMMs based on concave, increasing potential functions. This axiomatic approach argues that familiar CFMM designs emerge from desirable market properties, rather than merely being convenient formulas chosen independently of broader market-making principles.

• The central result establishes a two-way construction between suitable CFMMs on n assets and cost-function prediction markets with n possible outcomes. It connects functionality and information elicitation, showing that desirable conditions for facilitating asset trades correspond to desirable conditions for revealing beliefs through prediction-market activity.