logo

LMSR: How the Logarithmic Market Scoring Rule Works in Prediction Markets

LMSR: How the Logarithmic Market Scoring Rule Works in Prediction Markets
293

In a prediction market, the hard part is not only choosing the right outcome. The hard part is having a tradable price before a large crowd of traders arrives. LMSR solves part of that problem by creating automated liquidity and prices that move like probabilities.

LMSR stands for Logarithmic Market Scoring Rule. It is a mathematical mechanism used as an automated market maker in prediction markets. When traders buy or sell shares, LMSR adjusts the price of each outcome and limits the market maker's maximum loss.

This guide explains the formula, the b parameter, the link between liquidity and slippage, and the difference between LMSR and a traditional order book. The goal is technical but practical. You should come away understanding why this mechanism matters for new markets, especially when early liquidity is thin.

Risk note: this article is educational. It is not investment, trading, or platform advice. Prediction markets involve financial risk, resolution rules, variable liquidity, and possible loss of capital.

What Is LMSR?

LMSR is a logarithmic scoring rule designed to turn forecasts about future events into tradable prices. In prediction markets, it can work as an AMM for prediction-market outcomes. That means a trader does not need to wait for a direct counterparty to buy or sell.

The idea is associated with Robin Hanson's work on market scoring rules. The academic paper on Logarithmic Market Scoring Rules explains how scoring rules can aggregate information in markets. A PDF version is available from Harvard EECS.

A Simple Definition

Imagine a binary market: “Will the central bank keep rates unchanged at the next meeting?” There are two outcomes: Yes and No. If nobody wants to sell Yes when you want to buy it, a pure order book can stall. LMSR avoids that cold-start problem.

It always offers a price. When you buy Yes, the price of Yes rises. At the same time, the price of No falls. The market stays tradable even with a small number of participants.

Why It Is Called a Logarithmic Market Scoring Rule

The name comes from the logarithmic function used to calculate the total cost of the market state. That function connects three things: the number of shares already bought, the sensitivity of the price, and the sum of all outcome probabilities.

The “scoring rule” part comes from the academic origin. Scoring rules reward better forecasts. In markets, that logic becomes a structure where each trade updates a collective probability.

LMSR as an AMM for Prediction Markets

An automated market maker is a system that quotes prices with a formula instead of relying only on buyers and sellers being present at the same time.

In LMSR, the formula is not built to swap tokens like a constant-product AMM. It is built to price event outcomes. That makes LMSR specific to prediction markets and information aggregation.

How LMSR Calculates Prices

LMSR calculates prices through a cost function. The cost function tells the trader how much the market charges to move from the current state to a new state. The marginal price of each outcome comes from the slope of that function.

The Cost Formula: C(q) = b ln(Σ e^(qᵢ/b))

The standard formula is:

C(q) = b · ln(Σ e^(qᵢ / b))

Where:

  • qᵢ is the number of shares already issued for outcome i.
  • b is the liquidity parameter.
  • n is the number of possible outcomes.
  • Σ sums across all outcomes in the market.

In practice, the trader pays the difference between the cost after the order and the cost before the order. If the order changes the market state a lot, the cost rises. Larger orders therefore face more price impact.

The Price Formula: Marginal Probability for Each Outcome

The marginal price of each outcome is:

pᵢ = e^(qᵢ / b) / Σ e^(qⱼ / b)

This price sits between 0 and 1. In a binary market, Yes and No prices sum to 1 before possible fees or product adjustments. A price of 0.63 can be read as an implied probability near 63%.

That number is not a guarantee. It is the collective market price at that moment. It can change when new information arrives or when traders with capital move the market.

A 50/50 Example Before and After a Buy

Imagine a binary market starting neutral. No shares have been bought for Yes or No. With q_Yes = 0 and q_No = 0, both prices start near 50%.

Now a trader buys Yes shares. LMSR recalculates the cost and increases the marginal price of Yes. If the buy is small, Yes might rise from 50% to 53%. If the buy is larger, it could move to 60% or more.

The key point is that the trader does not receive one fixed price for the entire order. The first part of the order is cheaper. The final part costs more because the trade itself pushes the probability higher.

Situation Yes Price No Price Practical reading
Initial market 50% 50% Balanced uncertainty
Small Yes buy 53% 47% Mild pro-Yes signal
Larger Yes buy 60% 40% The order moved the market
Another large buy 70% 30% Slippage is more visible

The numbers are simplified. The actual move depends on the b parameter and order size. Still, the intuition is clear: LMSR prices are dynamic.

The Role of the b Parameter: Liquidity, Slippage, and Risk

The b parameter is the practical center of LMSR. It controls the algorithmic depth of the market. If b is low, the price moves quickly. If b is high, the price moves slowly.

Low b: More Sensitive Prices

With a low b, small orders can move the probability a lot. This can be useful for experimental markets or low-budget markets because the market maker commits less capital.

The tradeoff is higher slippage. A trader may see Yes at 50% but pay a higher average price if the order is meaningful. The market becomes more sensitive to isolated traders.

High b: More Depth and Less Slippage

With a high b, the market absorbs larger orders without a large price move. That improves the experience for traders who want to enter and exit with less impact.

But depth has a cost. The market maker accepts a larger maximum risk. Choosing b is therefore not a minor technical detail. It is a market-design decision.

Market Maker Maximum Loss: b ln(n)

One advantage of LMSR is bounded maximum loss. For a market with n outcomes, the theoretical maximum loss for the market maker is:

b · ln(n)

In a binary market, n = 2. The maximum loss is b · ln(2). As b rises, the market becomes more liquid, but the risk budget also rises.

This property matters for platforms. It allows them to seed liquidity without taking unlimited risk. Still, bounded does not mean trivial. The capital must be planned.

Practical Example: A Central Bank Decision Market

Now apply LMSR to a real type of event: a market asking whether a central bank will keep its policy rate unchanged at the next meeting. This kind of event has public information, a scheduled decision, and traders with different views.

Binary Market: Hold or Change Rates

At launch, the market might open at 50% for Hold and 50% for Change. Then new inflation data, central bank communication, and bank research come out. Traders react to that information.

If many traders buy Hold, LMSR increases the price of that outcome. If a more hawkish data point later pushes traders toward Change, the price can move back. The market becomes a live panel of expectations.

How One Order Moves the Probability

Suppose Hold is priced at 55%. A trader believes the fair price is 65%. The trader buys Hold shares. The order moves the price to 58%, then 60%, and maybe 62%, depending on size.

That move has two meanings. First, the trader expressed conviction with capital. Second, the next traders now face a different price. The market incorporated information or at least buying pressure.

What a Trader Should Check Before Trading

Before trading, a trader should check three things. Is the current price far from their own estimate? Will the order create too much price impact? Are the resolution rules clear?

It also helps to separate price from truth. A market at 62% does not know the future. It shows an implied probability created by trades, liquidity, and available information.

LMSR vs CLOB: Which Microstructure Makes More Sense?

LMSR is not the only way to run prediction markets. Many platforms use a central limit order book, or CLOB. Others use hybrid models.

A weak claim would be to say every popular platform uses LMSR. That is not safe. Polymarket and Kalshi, for example, are often discussed in relation to order-book dynamics or their own market structures. The useful point is to compare microstructures, not to make unsupported platform claims.

Order Books Depend on Counterparties

In a CLOB, buyers and sellers post prices and quantities. A trade happens when an order meets a compatible quote. This model works very well when there is enough liquidity.

The problem appears in new or small markets. If there is no counterparty, the trader may not execute. If the spread is wide, the displayed probability becomes less informative.

LMSR Provides Algorithmic Liquidity

LMSR solves the cold-start problem. It always quotes a price, even when few traders are active. That can be decisive for long-tail markets such as local events, technical questions, or niche economic releases.

The tradeoff is price impact. The liquidity comes from the formula, not from a deep queue of traders. If b is poorly calibrated, the market can become too sensitive or too expensive to maintain.

Why Platforms May Prefer a Hybrid Model

Hybrid models try to combine the advantages. An AMM can provide base liquidity. Then traders can add limit orders. The market gets early execution and improves efficiency as it grows.

The right choice depends on the product, user base, cost of capital, and operational complexity. There is no perfect microstructure for every case.

Criterion LMSR CLOB / Order Book Hybrid model
Initial liquidity Strong Weak without counterparties Strong
Efficiency in large markets Good, but depends on b Strong Strong
Slippage Controlled by b Controlled by real depth Mixed
Market-maker risk Bounded by b ln(n) Lower for the platform Depends on design
Best use New and long-tail markets Liquid markets Scaling platforms

Advantages and Limitations of LMSR

LMSR is elegant because it turns an academic rule into market infrastructure. But it does not remove risk. It does not create information from nothing, and it does not make every price correct.

Advantages: Continuous Liquidity, Coherent Prices, Bounded Loss

The first advantage is continuous liquidity. The trader always finds a quote. That reduces the empty-market problem.

The second advantage is coherence. In mutually exclusive markets, prices behave like probabilities that sum to 100% before fees and adjustments. That helps readers interpret the market.

The third advantage is bounded loss. The operator knows the maximum risk budget for the market. That predictability helps product design.

Limitations: b Calibration, Capital Cost, and Large-Market Incentives

The main limitation is b calibration. If b is too low, traders can move the market with little capital. If b is too high, the platform commits excessive capital.

There is also a capital cost. Automated liquidity looks simple on screen, but someone funds the risk. In large markets, a deep CLOB can be more efficient.

Finally, LMSR does not prevent strategic behavior. Traders can test depth, exploit low liquidity, or move prices in markets with little attention. Transparency and limits still matter.

The Risk of Treating Price as Absolute Truth

Price is a signal, not truth. A 70% price can be useful, but it still fails 30% of the time in an honest probabilistic reading. It can also reflect low liquidity, participant bias, or access restrictions.

Professional traders therefore use market probabilities as an input. They compare the price with their own models, official sources, public data, and execution risk.

Frequently Asked Questions About LMSR

Is LMSR the Same as an AMM?

LMSR can be used as a type of AMM for prediction markets. But not every AMM is LMSR. Constant-product AMMs, common in DeFi, follow a different logic. LMSR was designed for event-outcome prices.

Does LMSR Manipulate Probabilities?

Not in the usual sense. LMSR applies a public formula to update prices as trades happen. Still, large traders can move probabilities in low-liquidity markets. That is why price impact and the b parameter matter.

How Is LMSR Different From Sportsbooks?

In LMSR, the price comes from a transparent rule and changes with trading. In a traditional sportsbook, the house sets lines and manages exposure. The core difference is microstructure: prediction markets seek probability discovery, not only a quote set by the house.

Does LMSR Work for Markets With More Than Two Outcomes?

Yes. LMSR works for markets with multiple outcomes. The formula sums across all possible outcomes, and each marginal price represents that outcome's probability. Maximum loss grows with ln(n), where n is the number of outcomes.

What Is the Best Way to Study LMSR?

Start with the intuition: price rises when traders buy an outcome, and b controls the strength of that move. Then read practical materials such as the Cultivate Labs guide to LMSR. Finally, move into papers and simulations.

Conclusion: Why LMSR Matters in Prediction-Market Microstructure

LMSR matters because it solves a real market-design problem: new markets need liquidity before they have many traders. The Logarithmic Market Scoring Rule offers a transparent way to quote prices, update probabilities, and bound the market maker's maximum loss.

At the same time, LMSR requires care. The b parameter sets the balance between liquidity, slippage, and risk. Poor design can make a market too sensitive or too expensive. Good design creates a clearer environment for trading and analysis.

For traders, the practical lesson is simple. When you look at a market about central banks, elections, crypto, or sports, do not only read the displayed probability. Ask which microstructure produced that price, how much liquidity exists, and how much your order can move the market.

That is the difference between reading probability as a headline and reading probability like a trader. LMSR does not predict the future by itself. It organizes incentives so that information, capital, and risk can appear as price.

Author
Dale Hanson
Dale HansonCommunity Manager. X @InHansonWeTrust