Risk in Market Making

Inventory Risk

The number one risk in market making is inventory risk — the risk that the assets you are holding move against you while you wait to complete a round trip. When a market maker buys 1,000 shares at $50, they now have exposure to that stock. If the price drops to $49, they have an unrealized loss of $1,000 — regardless of how tight their spread was.

Inventory risk is the reason market makers cannot simply set a wide spread and collect money forever. The wider the spread, the fewer trades they execute, and the longer they hold inventory. The longer they hold, the more exposed they are to price movements. This tension between spread width and inventory duration is at the heart of optimal market-making theory.[1]

Adverse Selection

Adverse selection is the market maker’s most insidious enemy. It occurs when the counterparty to a trade possesses information that the market maker does not. An informed trader who knows a stock is about to drop will happily sell to a market maker at the current bid. The market maker, unaware of the impending decline, takes on a position that immediately becomes a loser.[2]

In crypto markets, adverse selection often manifests as latency arbitrage — traders with faster data feeds or lower-latency connections detect price changes on other exchanges before the market maker can update their quotes, and systematically trade against stale prices.

Volatility Risk

Volatility is a double-edged sword for market makers. On one hand, higher volatility means wider spreads and more revenue per trade. On the other hand, it means faster and larger price swings that can turn inventory positions into significant losses in seconds. During extreme market events — flash crashes, liquidation cascades, surprise news — quotes can become dangerously stale almost instantly.

Professional market makers use dynamic spread adjustment: as real-time volatility increases, spreads automatically widen. In extreme conditions, they may pull all quotes entirely (a “kill switch”) rather than risk catastrophic losses from quoting stale prices.

The "Picking Up Pennies" Problem

Market making is often described as “picking up pennies in front of a steamroller.” The attribution varies — it has been credited to everyone from Nassim Taleb to anonymous Wall Street traders — but the metaphor is apt. Market makers earn small, consistent profits on most trades. But occasionally, a large adverse move wipes out weeks or months of accumulated gains.

This return profile — many small wins, occasional large losses — creates a distribution that is negatively skewed. It can look deceptively profitable on a day-to-day basis while hiding fat-tail risks. Successful market makers obsess over managing these tail risks, because a single bad drawdown can undo months of careful work.

How Professionals Manage Risk

The most sophisticated market-making firms employ multiple layers of risk management:

• Dynamic spreads: automatically widen when volatility spikes or inventory grows.
• Position limits: hard caps on how much inventory can accumulate before quotes are pulled.
• Hedging: offsetting positions with correlated instruments or derivatives.
• Kill switches: automated systems that halt trading entirely when losses exceed predefined thresholds.
• Skewing: adjusting bid and ask prices asymmetrically to encourage trades that reduce inventory.

The Avellaneda-Stoikov Framework

In 2008, Marco Avellaneda and Sasha Stoikov published what has become the foundational paper on optimal market making under inventory risk. Their framework provides a mathematically rigorous way to compute the “reservation price” — a fair price adjusted for current inventory — and the optimal spread to quote around it.[1]

The key insight is elegant: a market maker holding excess inventory should shift their quotes to encourage trades that reduce that inventory. If you hold too much of an asset, lower your ask price to attract buyers. If you are short, raise your bid to attract sellers. The optimal amount of skewing depends on volatility, risk aversion, and how much time remains in the trading horizon.

Subsequent work by Guéant, Lehalle, and Fernandez-Tapia extended this framework to account for multiple assets, execution uncertainty, and more realistic market conditions.[2] This line of research has become the theoretical backbone of modern algorithmic market making, and forms the basis of many institutional strategies operating today.[3]