Market Making on Prediction Markets: Real-World Case Study

# Market Making on Prediction Markets: Real-World Case Study **Market making on prediction markets means quoting both a buy and sell price simultaneously, collecting the spread as profit while providing liquidity to other traders.** In practice, a market maker on platforms like Polymarket or Kalshi posts YES shares at $0.55 and NO shares at $0.45 simultaneously on a 50/50 event, pocketing the difference on each matched trade. This guide walks through real-world examples, actual profit mechanics, and the risks that can wipe out even well-capitalized operations overnight. --- ## What Is Market Making in Prediction Markets? **Market making** is the practice of continuously quoting both sides of a market — a bid (buy price) and an ask (sell price) — so other participants always have someone to trade against. On traditional exchanges, firms like Citadel and Virtu do this at scale. On **prediction markets**, the concept translates directly, but with a twist: the underlying asset isn't a stock or commodity — it's a probability. When you trade on a platform like Polymarket, you're buying YES or NO shares in a binary outcome. A YES share pays $1 if the event happens, $0 if it doesn't. The market maker sits in between, quoting: - **Bid**: The highest price they'll pay to buy YES shares from you - **Ask**: The lowest price they'll sell YES shares to you The difference is the **bid-ask spread**, and that's where market maker revenue comes from. On decentralized platforms using **Automated Market Maker (AMM)** models (like early Augur versions), liquidity providers deposit into pools and earn fees proportionally. On **Central Limit Order Book (CLOB)** platforms like Polymarket's current architecture and Kalshi, market makers actively manage orders. --- ## Real-World Case Study #1: The 2024 U.S. Presidential Election The 2024 U.S. presidential election on Polymarket was the largest-volume prediction market event in history, trading over **$3.7 billion** in total volume. This created an extraordinary environment for market makers. ### How the Spread Moved Over Time | Time Period | Typical Bid-Ask Spread (Trump WIN) | Daily Volume | |---|---|---| | 6 months before election | $0.06 – $0.08 | $2M–$5M | | 1 month before election | $0.02 – $0.04 | $30M–$80M | | Election week | $0.005 – $0.015 | $200M+ | | Election night | $0.001 – $0.003 | $500M+ | As volume surged, spreads compressed dramatically. Early in the cycle, a market maker quoting a $0.06 spread on a $10,000 position could collect roughly **$300 per round-trip** (if both sides filled). By election night, that same position earned just $15–$30 in spread — but turnover was 100x higher. **Sophisticated market makers** rotated capital quickly, using automated bots to update quotes every few seconds as new polls, news events, and betting flows moved the consensus probability. Platforms like [PredictEngine](/) enable exactly this kind of rapid automated quoting. ### The Inventory Risk Problem One well-documented failure mode: a market maker built up a heavily skewed **YES inventory** on Harris winning after a major polling update in October 2024 suggested a late surge. They held roughly 200,000 YES shares at an average cost of $0.47. When the actual outcome became clear on election night, those shares moved toward $0, creating an unrealized loss of approximately **$94,000** before they could unwind. This is called **inventory risk** — the danger that your market-making position becomes directionally exposed when you can't rebalance fast enough. --- ## Real-World Case Study #2: NBA Playoffs Market Making **Sports markets** offer unique market-making opportunities because they're time-bounded, heavily researched, and driven by verifiable real-world events. The NBA Playoffs generate consistent liquidity across dozens of simultaneous markets. Consider a market maker operating during the 2024 NBA Finals on Polymarket. The "Boston Celtics to win the 2024 NBA Championship" market saw sustained volume throughout the playoffs. ### A Day in the Life of a Sports Market Maker Here's how a typical day unfolded for an active market maker during a Game 5: 1. **Pre-game**: Quote at $0.64/$0.60 (Celtics YES/NO) with $500 resting on each side 2. **Tip-off announcement of key player injury**: Prices shift suddenly; market maker cancels orders within 200ms using an automated system 3. **Re-price**: New quotes set at $0.55/$0.51 reflecting updated probability 4. **First quarter**: Gradual fills on both sides, net $180 collected in spread 5. **Halftime**: Celtics up 12; YES price moves to $0.78; market maker refreshes quotes 6. **Final buzzer**: Celtics win; YES resolves at $1.00; net position was slightly long, gained $240 extra on residual inventory **Total daily P&L from spread alone**: approximately $420 on ~$15,000 in deployed capital — a 2.8% single-day return. Annualized, consistent performance at this rate would be extraordinary, though individual days vary wildly. For deeper strategy on sports prediction market trading, see this breakdown of [AI-Powered Polymarket Trading During NBA Playoffs](/blog/ai-powered-polymarket-trading-during-nba-playoffs), which covers how automated tools are redefining the edge in sports markets. --- ## How to Set Up a Basic Market Making Operation Getting started requires more than just capital. Here's a structured approach: 1. **Choose your platform carefully**: Polymarket (CLOB, crypto-based) vs. Kalshi (regulated, USD) have different fee structures and market depth. Review the [trader playbook comparing Polymarket vs Kalshi with limit orders](/blog/trader-playbook-polymarket-vs-kalshi-with-limit-orders) before committing capital. 2. **Complete onboarding and KYC**: For regulated platforms like Kalshi, identity verification is mandatory. Even on Polymarket, wallet setup and risk controls matter significantly — see this [KYC and wallet setup risk analysis](/blog/kyc-wallet-setup-risk-analysis-for-ai-prediction-markets) before depositing. 3. **Fund your account with appropriate capital**: Most serious market makers start with a minimum of **$5,000–$25,000** to have meaningful resting order sizes. Too little and your fills are inconsistent. 4. **Build or subscribe to a quoting system**: Manual quoting is nearly impossible at scale. Automated systems update quotes based on external signals (news, odds from other platforms, social sentiment). 5. **Set hard inventory limits**: Define maximum exposure in any single market direction. A common rule: no more than **15% of capital** in one-sided inventory at any time. 6. **Monitor and adjust spreads dynamically**: Wider spreads during low information periods, tighter spreads in high-volume markets where competition from other makers is fierce. 7. **Track fees carefully**: Polymarket charges **2% of winnings**; Kalshi charges between **1%–7%** depending on market type. These eat directly into spread revenue. For those with limited budgets, you don't need six figures to start — check out this guide on [AI-powered prediction market order book analysis on a small budget](/blog/ai-powered-prediction-market-order-book-analysis-on-a-small-budget) to understand what's achievable with smaller deployments. --- ## The Economics of Spread Capture vs. Adverse Selection The fundamental tension in market making is simple: **you want to fill orders from uninformed traders, not informed ones.** An uninformed trader sells YES shares because they need liquidity — they want to exit their position regardless of the underlying probability. That's a profitable fill for you. An **informed trader** sells YES shares because they know something you don't — perhaps a news event is imminent that will crash the probability. Filling their order means you've just bought into a losing position. This is called **adverse selection**, and it's the silent killer of market making operations. A 2023 academic study on decentralized prediction markets found that market makers faced adverse selection costs of approximately **18–24% of gross spread revenue** during high-information events (elections, major legal rulings, central bank decisions). ### Strategies to Reduce Adverse Selection | Strategy | Description | Effectiveness | |---|---|---| | Widen spreads before known events | Earn more per fill to cover potential losses | High | | Use time-based order cancellation | Pull quotes automatically near news releases | High | | Monitor order flow imbalance | Detect when informed traders are active | Medium | | Quote smaller sizes | Limit exposure per fill during uncertain periods | Medium | | Focus on high-volume markets | More uninformed flow dilutes informed trading | Medium-High | --- ## Real-World Case Study #3: Corporate Earnings Prediction Markets **Earnings markets** on platforms that allow them (primarily Kalshi and some offshore venues) are a fascinating case because they combine quantitative modeling with market-making mechanics. Take NVDA earnings markets as an example. Before a major Nvidia earnings announcement, a market maker might see markets like "Will NVDA beat EPS estimates by more than 10%?" trading at $0.40/$0.37 (bid/ask). Volume is moderate pre-announcement, but spikes massively in the **2-hour window before the report drops**. The smart play: **widen spreads significantly** — to $0.08 or more — in the 30 minutes before the announcement, then pull all quotes entirely 5 minutes before. After the announcement, re-enter with tight spreads as price discovery completes. This framework allows collection of wide spreads during uncertainty while avoiding catastrophic adverse selection at the moment of information release. For a detailed look at how institutional players approach these markets, the [NVDA earnings risk analysis for institutional investors](/blog/nvda-earnings-risk-analysis-what-institutional-investors-need) provides a useful framework. --- ## Automated Market Making: Tools and Infrastructure Professional market makers don't sit at a desk refreshing screens. They run **automated systems** that: - Pull real-time odds from correlated markets (sports books, financial derivatives, other prediction platforms) - Apply pricing models to determine fair value - Submit, modify, and cancel orders via API in milliseconds - Monitor inventory and enforce risk limits automatically The gap between manual and automated market making is enormous. A manual maker might update quotes 50 times per day; an automated system might update **50,000 times per day**. On a high-volume election market, that difference in responsiveness is the difference between profitability and getting picked off. [PredictEngine](/) provides infrastructure purpose-built for this kind of systematic trading, including order management, analytics, and market connectivity across major prediction platforms. --- ## Risks That Can Destroy a Market Making Operation Despite the appeal of "earning the spread," market making carries serious risks that beginners underestimate: - **Event risk**: A sudden, unexpected outcome (assassination, natural disaster, regulatory ruling) can instantly move prices 30–50 points, leaving market makers with catastrophic inventory losses - **Liquidity risk**: In thin markets, you might not be able to exit an inventory position at any reasonable price - **Counterparty risk**: On some platforms, smart contract bugs or platform insolvency can freeze funds - **Fee drag**: As spreads compress, fees can consume 30–50% of gross spread revenue in competitive markets - **Regulatory risk**: Prediction market regulation is evolving rapidly; markets can be shut down overnight Understanding the full economic picture before deploying capital is essential. The [economics of prediction markets quick reference guide](/blog/economics-prediction-markets-quick-reference-step-by-step) covers these fundamentals in accessible terms. --- ## Frequently Asked Questions ## How much capital do I need to start market making on prediction markets? Most active market makers start with **$5,000 to $25,000** in available capital to maintain meaningful order sizes and survive inventory drawdowns. Below $1,000, it's difficult to quote sizes large enough to earn meaningful spread revenue after fees. ## What is the typical profit margin for prediction market makers? Gross spread margins vary widely — from **0.5% to 8%** of notional depending on market liquidity and competition. After adverse selection and fees, net margins of **1–3% monthly** on deployed capital are considered solid for consistent operators. ## Is market making on prediction markets legal? In most jurisdictions, trading on regulated platforms like **Kalshi** is fully legal for U.S. residents. Polymarket is decentralized and accessible globally, though U.S. residents face restrictions. Always consult local regulations and platform terms before trading. ## How do automated market makers (AMMs) differ from manual market making? **AMMs** use algorithmic formulas (like constant product models) to automatically set prices based on pool ratios, removing the need for active quote management. Manual and algorithmic CLOB market makers actively post and update orders to express specific pricing views and earn spreads more precisely. ## What markets are best for beginner market makers? High-volume, low-volatility markets with predictable resolution timelines are best for beginners — think recurring sports leagues or economic indicator markets rather than political events. Higher volume means more uninformed flow and less adverse selection risk. ## Can AI tools improve market making performance on prediction markets? Yes — AI tools can analyze order book depth, detect flow imbalances, and update quotes in real time far faster than humans. Platforms like [PredictEngine](/) combine market data feeds with automated quoting tools specifically designed for prediction market environments. --- ## Start Market Making Smarter With PredictEngine Market making on prediction markets is one of the few genuine **edge strategies** available to retail and semi-professional traders today — but execution matters enormously. The difference between profitable and unprofitable market making often comes down to speed, pricing accuracy, and disciplined risk management. [PredictEngine](/) gives you the tools to compete: real-time order book data, automated quoting infrastructure, and analytics built specifically for prediction markets. Whether you're a first-time liquidity provider or a systematic trader scaling an existing operation, PredictEngine's platform is designed to help you capture spread efficiently while managing the risks that trip up less-prepared participants. **Start your free trial today** and see how professional-grade market making tools can transform your approach to prediction market trading.