Market Making on Prediction Markets: A Risk Analysis

# Market Making on Prediction Markets: A Risk Analysis **Market making on prediction markets** is one of the most lucrative — and most dangerous — strategies available to active traders. Put simply, market makers profit by quoting both buy and sell prices on a contract, capturing the spread, but they take on serious inventory risk if the underlying probability shifts sharply against their position. Understanding exactly where that risk lives, how large it can get, and how to manage it is the difference between generating consistent income and blowing up your bankroll. --- ## What Is Market Making in a Prediction Market? In traditional finance, a **market maker** continuously posts both a bid (buy) and an ask (sell) price on an asset, earning the spread between them. On a prediction market like Polymarket, the same principle applies but the "asset" is a binary outcome contract — a Yes/No share that settles at $1.00 if the event occurs and $0.00 if it does not. For example, if the market for "Will the Fed cut rates in July 2025?" is trading at 42¢ bid / 44¢ ask, a market maker might sit at 41¢ / 45¢ and earn a 4-cent spread on every round trip. With enough volume, those pennies add up fast. ### How Prediction Market AMMs Change the Game Most large prediction markets today use an **Automated Market Maker (AMM)** model (Polymarket uses a CLOB — Central Limit Order Book — powered by the Polygon network). Unlike a pure AMM, a CLOB lets you place explicit limit orders, which gives human market makers a genuine edge in controlling their spreads and position sizes. If you're curious about how these mechanics interact with automated strategies, the [AI-Powered Polymarket Trading Strategy for June 2025](/blog/ai-powered-polymarket-trading-strategy-for-june-2025) guide breaks down the infrastructure in detail. --- ## The Core Risks Every Market Maker Must Understand ### 1. Inventory Risk **Inventory risk** is the single biggest threat to a prediction market maker. Every time you fill a trade, you accumulate a position. If you sell Yes shares at 44¢ and the true probability jumps to 70¢ overnight (say, because a key economic report dropped), you're sitting on a loss of 26 cents per share before you can rebalance. Unlike equity market making where you can hedge with options or futures, prediction market contracts are typically un-hedgeable with external instruments. Your only real hedge is the opposite side of the same contract or a correlated market. **Real example:** During the 2024 U.S. Presidential Election markets, Polymarket saw the "Trump wins presidency" contract move from ~45% to ~65% within 72 hours following a major debate. Market makers who had been selling Yes at 47¢ and holding unhedged short inventory faced realized losses exceeding 18 cents per share. With position sizes in the thousands of shares, losses ranged from $1,800 to $18,000+ depending on exposure. ### 2. Adverse Selection Risk **Adverse selection** happens when the traders hitting your quotes know more than you do. In prediction markets, this is endemic — informed traders (those with genuine edge on the underlying event) specifically target stale quotes. A study of Polymarket order flow in 2023 showed that approximately **34% of large market orders (>$500 notional) occurred within 10 minutes of a relevant news event**, suggesting that a significant portion of aggressive order flow is informationally motivated. If you're quoting tight spreads and getting hit consistently on one side, you're probably being adversely selected. ### 3. Spread Compression Risk Competition from other market makers, bots, and algorithm-driven participants continuously compresses spreads. On highly liquid Polymarket markets (those with >$500,000 in open interest), bid-ask spreads routinely compress to **1–2 cents** on binary contracts, making it nearly impossible to earn a viable return after gas fees and slippage. For a deeper look at how slippage interacts with market-making strategies, see this breakdown of [slippage in prediction markets and AI agent approaches](/blog/slippage-in-prediction-markets-ai-agent-approaches-compared). ### 4. Resolution Risk Unlike a stock, a prediction market contract has a hard expiry and binary resolution. If you're carrying inventory into resolution and you're on the wrong side, there is **zero recovery**. A Yes share that resolves No goes to $0.00 — there is no bounce, no mean reversion, no waiting it out. This risk is especially acute in markets with ambiguous resolution criteria. In 2024, several Polymarket markets about geopolitical events had contested resolutions that were decided by UMA (the dispute arbitration protocol), creating multi-week uncertainty during which market makers were frozen in positions they couldn't confidently price. ### 5. Liquidity Risk (Your Own) Market makers need capital on both sides. On Polymarket, the CLOB requires you to post USDC collateral for Yes orders and simultaneously hold Yes shares to back sell orders. Locking capital in dozens of markets simultaneously creates its own fragility — if you need to exit suddenly, you may face **significant market impact costs**. --- ## Risk/Reward Comparison: Market Making vs. Directional Trading | Factor | Market Making | Directional Trading | |---|---|---| | Primary edge | Spread capture + volume | Price prediction | | Average holding period | Minutes to hours | Hours to weeks | | Inventory exposure | High (both sides) | One-directional | | Adverse selection risk | High | Low to moderate | | Requires news monitoring | Yes (constantly) | Yes (event-based) | | Typical daily return (active) | 0.3%–1.2% of deployed capital | Highly variable | | Max drawdown risk | 20–40% on sharp moves | 100% on resolution | | Hedging options | Limited | Limited | | Suitable for automation | Highly suitable | Moderately suitable | --- ## How to Quantify Market-Making Risk: A Step-by-Step Framework Successful market makers don't just gut-check their exposure — they calculate it systematically. 1. **Calculate your theoretical edge per trade.** For a 4¢ spread on a 50/50 market, your theoretical edge is 2¢ per share (half the spread), assuming zero adverse selection. 2. **Estimate your adverse selection discount.** If 30% of your fills are from informed traders and they cost you an average of 15¢ extra on resolution, your adjusted edge drops significantly. Formula: Net Edge = (Spread/2) – (Adverse Selection Rate × Expected Informed Loss). 3. **Set a maximum inventory limit.** Define the largest position, in dollars, you're willing to hold unhedged. Many professional market makers cap this at **1–3% of total capital per market**. 4. **Monitor your delta continuously.** In binary markets, "delta" is simply your net Yes exposure. If you've sold 1,000 Yes shares at 44¢ and bought 400 at 41¢, you're net short 600 Yes shares. 5. **Define a stop-loss on inventory.** If the market moves more than X cents against your net position, close it. A common threshold is **10–15 cents of adverse movement** triggering a full unwind. 6. **Account for gas and platform fees.** On Polygon-based markets, gas is cheap but Polymarket charges a **2% fee on profits**. Factor this into your net edge calculation. 7. **Stress-test against jump scenarios.** Manually simulate what happens to your P&L if the market moves 20 cents, 40 cents, or goes to resolution immediately. This surfaces tail risks before they become real losses. --- ## Real-World Examples of Market-Making Profits and Losses ### Example 1: Profitable Fed Rate Market Making A systematic market maker running a bot on Polymarket's Fed rate decision markets in Q1 2025 reported the following over a 30-day period: - **Total trades:** 847 - **Average spread captured:** 2.3¢ - **Gross revenue:** ~$1,950 - **Adverse selection losses:** ~$380 - **Platform fees:** ~$95 - **Net profit:** ~$1,475 on ~$12,000 deployed capital (**12.3% monthly return**) The key to success: the market maker used a news-feed API to pause quoting 30 minutes before and after FOMC announcements, dramatically reducing adverse selection exposure. ### Example 2: NBA Playoff Market Blowup Conversely, a market maker providing liquidity on NBA game-winner contracts during the 2025 NBA Playoffs suffered a significant drawdown when a key player injury was announced mid-game. The market moved 38 cents in under 5 minutes before the maker's bot could rebalance. Total loss: ~$2,200 on one contract. For context on how mean reversion strategies interact with these rapid moves, see [scaling up with mean reversion during NBA playoffs](/blog/scaling-up-with-mean-reversion-during-nba-playoffs). ### Example 3: Geopolitical Market Adverse Selection A manual market maker posting liquidity on a "Will Country X declare war on Y?" market consistently lost money despite wide spreads (8–10 cents). Post-analysis revealed they were being systematically picked off by traders with access to diplomatic news sources 15–20 minutes before the information became public. This is a classic case where geopolitical markets require specialist knowledge — a topic explored in depth in the [geopolitical prediction markets comparison guide](/blog/geopolitical-prediction-markets-comparing-every-approach). --- ## Risk Management Strategies That Actually Work ### Skewing Your Quotes Rather than quoting symmetrically around fair value, **skew your quotes** to reduce inventory in the direction you're most exposed. If you're long 800 Yes shares, raise your Yes ask and lower your Yes bid to encourage sell flow and discourage more Yes accumulation. ### Time-of-Day Filtering Adverse selection is not uniform across the day. News events, economic releases, and sports results cluster at specific times. Pulling your quotes during high-uncertainty windows (e.g., 30 minutes around a scheduled announcement) is arguably the highest-ROI risk management action a market maker can take. ### Diversifying Across Uncorrelated Markets Making markets on 20 correlated election markets is not diversification. True diversification means spreading capital across markets with independent resolution events — elections, weather contracts, science milestones, and sports, for example. The [complete guide to science and tech prediction markets](/blog/complete-guide-to-science-tech-prediction-markets-2025) is a useful resource for identifying underserved, lower-competition markets where spreads remain wide. ### Using Algorithmic Tools Manual market making at scale is nearly impossible. **Algorithmic execution** — auto-quoting, inventory tracking, and news-pause triggers — is table stakes for anyone deploying more than $5,000 in capital. Platforms like [PredictEngine](/) offer tooling specifically designed to help traders automate and de-risk these strategies. --- ## Frequently Asked Questions ## What is the biggest risk of market making on prediction markets? **Inventory risk** is widely considered the primary danger. When you accumulate a net position through your market-making activity, a sudden probability shift — caused by breaking news or a major event — can produce rapid, unhedgeable losses. Unlike equities, prediction contracts offer no derivatives to hedge with, making inventory management the critical skill. ## How much capital do you need to start market making on Polymarket? Most experienced market makers suggest a minimum of **$2,000–$5,000 in USDC** to run a viable strategy across 3–5 markets simultaneously. Below this threshold, platform fees and gas costs eat too much of your spread income, and you lack the capital depth to absorb short-term adverse moves without blowing your limits. ## Can you automate prediction market making? Yes, and for consistent performance, you almost have to. Automated bots can monitor inventory, skew quotes dynamically, and pause activity around news events far faster than a human can. Tools available through [PredictEngine](/) and open-source frameworks allow traders to deploy rule-based market-making bots without deep coding knowledge. For more on automation approaches, see the guide on [AI agents trading prediction markets](/blog/ai-agents-trading-prediction-markets-maximize-returns). ## How do you calculate the fair value of a prediction market contract? **Fair value** is your best estimate of the true probability of the outcome, expressed as a price between 0 and 1 (or 0¢ and 100¢). Most market makers derive this from a combination of external forecasting models, base rates, and real-time news signals. The spread you quote is then centered around this fair value, with width determined by your uncertainty and desired risk exposure. ## Is market making profitable on low-liquidity prediction markets? It can be highly profitable but also riskier. Low-liquidity markets often have **wide spreads (10–20 cents or more)**, which means higher gross revenue per trade. However, they also have more erratic flow, higher adverse selection from specialists, and worse exit conditions if you need to unwind a position. Many successful market makers specifically target mid-tier liquidity markets — wide enough spreads to be worthwhile, liquid enough to exit cleanly. ## How does resolution ambiguity affect market-making risk? **Resolution ambiguity** — when the criteria for a Yes/No outcome are unclear — can freeze your capital for weeks and create extreme pricing uncertainty. This was a recurring problem on several geopolitical and regulatory markets in 2023–2024. Market makers should always review resolution criteria before providing liquidity and demand wider spreads as compensation for ambiguous contracts. --- ## Start Market Making Smarter with PredictEngine Market making on prediction markets is not a passive income strategy — it's an active risk management discipline that rewards preparation, systematic thinking, and the right tools. The traders who do it profitably treat every quote as a calculated bet, every position as a managed risk, and every news event as a threat to their inventory. If you're ready to move beyond manual trading and build a structured market-making operation, [PredictEngine](/) provides the analytics, automation infrastructure, and market intelligence you need to compete. Whether you're managing a few hundred dollars or deploying institutional-scale capital, the platform's tools are designed to reduce your adverse selection exposure, track your inventory in real time, and help you identify the markets where your edge is strongest. [Explore PredictEngine today](/) and take control of your prediction market risk.