Market Making on Prediction Markets: Power User's Guide

# Market Making on Prediction Markets: Power User's Quick Reference **Market making on prediction markets means continuously posting both buy and sell orders to earn the bid-ask spread while managing your exposure to directional risk.** Done well, it's one of the most consistent edge strategies available to sophisticated traders — providing liquidity to others while capturing small, repeatable profits on each completed round-trip. This guide distills everything a power user needs: spread mechanics, inventory management, automation frameworks, and the key differences between platforms. --- ## Why Market Making on Prediction Markets Is Different Traditional equity market making happens in centralized, regulated venues with tight rules, co-location advantages, and professional competitors running sub-millisecond execution. **Prediction market making is still the wild west by comparison** — and that's a feature, not a bug. Platforms like Polymarket, Kalshi, and Manifold operate with thinner liquidity pools, slower price discovery, and a more diverse mix of participants — from informed traders to opinion-driven retail bettors. That mix creates **wider natural spreads** than you'd ever find in equities or crypto spot markets, which means market makers can extract real edge without needing nanosecond infrastructure. The key insight: on a well-chosen prediction market, a market maker can earn **2–8% per round-trip on low-volatility contracts** while staying genuinely delta-neutral. Compare that to stock market making, where typical spreads are measured in basis points. --- ## Core Mechanics: Spreads, Ticks, and Position Sizing ### The Bid-Ask Spread Your bread and butter is the **bid-ask spread** — the difference between the price you're willing to buy (bid) and the price you're willing to sell (ask). If a contract is trading around 45¢, you might post: - **Bid:** 43¢ - **Ask:** 47¢ - **Spread captured:** 4¢ per share, per completed round-trip The goal is for both sides to fill over time. If the market drifts to 50¢ and only your ask got hit, you've got inventory risk — you're now short a contract that moved against you. ### Position Sizing Rules of Thumb | Scenario | Recommended Max Exposure | |---|---| | Low-volatility binary (election, scheduled event) | 5–10% of capital per market | | High-volatility binary (breaking news, crypto price) | 1–3% of capital per market | | Correlated markets (same underlying event) | Treat as one position, combined limit | | Long-tail, illiquid contracts | 0.5–1% of capital, wider spreads | **Never size your inventory so large that a full adverse move would hurt your overall book by more than 2–3%.** Most blowups in prediction market making come from under-sizing spreads on high-vol contracts, not from bad fundamental views. --- ## Inventory Risk: The Market Maker's Biggest Enemy Inventory risk is what happens when your book becomes **directionally skewed** because one side of your market keeps getting hit. If everyone is buying YES from you on a political contract, you're accumulating a short position that needs to be hedged or offloaded. ### Managing Skew in Practice 1. **Widen the spread dynamically.** If you're long-heavy on a contract, raise your ask and lower your bid to discourage further buys and attract sellers. 2. **Set hard inventory limits.** Define a maximum net position per contract (e.g., ±200 shares). When you hit it, pull your quotes on the unfavorable side entirely. 3. **Cross-market hedging.** If Polymarket and Kalshi both list the same event, use one platform to offset inventory built on the other. This is a form of [prediction market arbitrage](/polymarket-arbitrage) that also serves a risk management function. 4. **Lean into correlated markets.** A large YES position on "Fed rate cut in September" might be partially offset by a YES position on "10-year yield above 4.5% in October" if those outcomes tend to move together. 5. **Time-decay to your advantage.** Many binary contracts lose uncertainty (and volatility) as the resolution date approaches. If you're stuck with inventory, a tightening spread environment may bail you out naturally. --- ## Spread Calibration: How Wide Should You Go? Getting your spreads right is a mix of art and data. **Too tight and you're giving away edge; too wide and you never fill.** Here's a practical calibration framework: ### Step-by-Step Spread Setting Process 1. **Observe the natural spread.** Before posting anything, watch the order book for 5–10 minutes to understand where natural liquidity clusters. 2. **Estimate implied volatility.** How fast has this market moved in the last 24 hours? Calculate a rough hourly standard deviation from recent trade history. 3. **Set your half-spread ≥ hourly IV.** If the market has been moving ±1¢/hour, your half-spread should be at least 1.5–2¢ to protect against adverse selection. 4. **Adjust for time to resolution.** Contracts expiring in 48 hours have concentrated risk — widen aggressively (3–4x your normal spread) as resolution approaches. 5. **Layer your orders.** Don't post a single large order at one price. Post smaller tranches at 43¢, 42¢, and 41¢ on the bid side. This reduces adverse selection and captures more of the order flow range. 6. **Monitor fill ratios.** If >80% of your fills are on one side, your quotes are skewed. Recalibrate immediately. 7. **Log everything.** Build a spreadsheet or database of every market, spread, fill ratio, and P&L. Pattern recognition over 100+ markets will teach you more than any model. For an example of how data-driven approaches pay off in real conditions, see this [Kalshi trading case study with real Q2 2026 results](/blog/kalshi-trading-case-study-real-results-for-q2-2026) — the fill ratio analysis alone is worth studying. --- ## Automation: Bots, APIs, and Execution Infrastructure Manual market making is exhausting and doesn't scale. **The power user's real edge is automation** — building or deploying bots that continuously reprice, manage inventory, and execute without human intervention. ### Minimum Viable Bot Architecture A functional prediction market making bot needs five components: - **Market data ingestion:** WebSocket feeds or polling the platform API for real-time order book state. - **Pricing engine:** Logic that computes your bid/ask based on current mid-price, inventory, and spread parameters. - **Order management system (OMS):** Handles placing, canceling, and replacing orders. Crucial for avoiding "stuck" orders that accumulate unwanted inventory. - **Risk module:** Hard limits on position size, daily loss limits, and circuit breakers that halt activity if the market behaves anomalously. - **Logging and alerting:** Every trade logged, P&L tracked in real time, and alerts for large inventory buildups or unusual market conditions. Platforms like Polymarket expose REST and WebSocket APIs that allow full programmatic trading. Kalshi similarly offers an API tier for institutional and power users. For a deeper dive into bot-powered approaches, the [AI-powered earnings surprise markets guide](/blog/ai-powered-earnings-surprise-markets-the-power-users-edge) covers how automated systems handle similar high-frequency decisions. You can also explore an [AI trading bot](/ai-trading-bot) or [Polymarket bot](/polymarket-bot) solution to accelerate your automation stack rather than building from scratch. ### LLM Integration for Dynamic Repricing Advanced setups are now incorporating **large language models** to inform repricing decisions — not for execution, but for context. An LLM can parse news headlines and flag when a market might be about to reprice sharply, giving your bot enough warning to widen spreads or pull quotes before getting adversely selected. For a full breakdown of approaches, [comparing LLM-powered trade signals](/blog/llm-powered-trade-signals-comparing-every-approach) is an essential read. --- ## Platform Comparison: Where to Market Make Not all prediction markets are created equal for liquidity providers. Here's how the major venues stack up: | Platform | Order Book Type | API Access | Typical Spreads | Best For | |---|---|---|---|---| | Polymarket | CLOB (Central Limit Order Book) | Yes (REST + WS) | 1–5¢ | Political, crypto, sports | | Kalshi | CLOB | Yes (institutional tier) | 2–8¢ | Macro, economic, weather | | Manifold | AMM (Automated Market Maker) | Limited | Variable | Niche, community markets | | Metaculus | Aggregation, no trading | No | N/A | Research only | | PredictIt | CLOB | No official API | 3–10¢ | US political (capped positions) | **Polymarket and Kalshi are the primary venues** for serious market making because they use central limit order books where your limit orders interact directly with incoming market orders. AMM-based platforms like Manifold work differently — your "liquidity provision" happens by trading against the AMM curve, which has different risk characteristics. For specific Kalshi strategies after major political events, see this [quick reference on Kalshi trading after the 2026 midterms](/blog/kalshi-trading-after-the-2026-midterms-quick-reference). --- ## Risk Controls Every Power User Needs Market making looks low-risk until it isn't. The following hard rules should be non-negotiable: - **Daily loss limit:** If you're down more than X% in a single day, halt all activity and review. X should be 3–5% of total deployed capital. - **Correlation limits:** Don't be a liquidity provider across 10 markets that all resolve YES/NO based on the same underlying event (e.g., multiple election-related contracts). Your "diversified" book can be 100% correlated. - **Resolution spike protection:** In the 2–4 hours before a major resolution, markets often exhibit dramatic price swings driven by breaking information. Many experienced market makers **go dark** during this window entirely. - **Adverse selection monitoring:** Track the ratio of informed vs. uninformed flow. A sudden surge of one-sided order flow is often the signal that someone knows something you don't. This is your cue to pull quotes. - **Regulatory awareness:** Different platforms have different legal classifications. The [tax guide for AI agents in prediction markets](/blog/tax-guide-ai-agents-in-weather-prediction-markets) covers some of the compliance considerations that apply equally to human and automated market makers. --- ## Advanced Strategies for Experienced Market Makers ### Quote Stuffing Defense Some sophisticated participants will send rapid sequences of small orders to test your bot's response time or trigger your repricing logic. Build **rate limiting and minimum resting time** into your OMS so you're not reacting to noise. ### Volatility Harvesting On markets with high implied volatility but slow actual resolution, you can earn more by **actively gamma scalping** — adjusting your net position in small increments as the price oscillates, while keeping your market making quotes live. This is an advanced overlay on top of standard spread income. ### Multi-Market Basket Making Rather than making markets one contract at a time, consider running a **basket of correlated contracts simultaneously**, netting your inventory across the full basket. This is particularly effective for [geopolitical prediction markets](/blog/advanced-geopolitical-prediction-markets-strategy-june-2025) where related events (same election, same policy decision) create natural hedging opportunities. --- ## Frequently Asked Questions ## What is the minimum capital needed to market make on prediction markets? **There's no hard minimum, but $5,000–$10,000 gives you enough to diversify across 10–20 markets meaningfully.** Below that, transaction costs and position minimums make it difficult to build a well-sized book. With $50,000+, you can run a systematized operation with measurable edge. ## How do I avoid getting picked off by informed traders? Watch for **one-sided order flow**, sudden volume spikes, and correlations with breaking news. Build adverse selection metrics into your monitoring — if your fill ratio on one side exceeds 70% over a short window, widen or pull quotes immediately. ## Can I market make on both Polymarket and Kalshi at the same time? **Yes, and this is actually a best practice** — running correlated books on both platforms lets you use one to hedge the other, effectively arbitraging the spread between venues while maintaining liquidity provision on both. See [prediction market arbitrage strategies](/polymarket-arbitrage) for technical implementation details. ## How does time to expiration affect my spread strategy? **The closer to resolution, the wider your spreads should be.** In the final 24–48 hours, a single piece of information can move a contract from 45¢ to 95¢ in minutes. Most experienced market makers widen to 10–20¢ spreads (or exit entirely) within 6 hours of resolution. ## Are there automated tools that handle prediction market making for me? **Yes.** Platforms like [PredictEngine](/) offer infrastructure and tooling designed for automated prediction market strategies, including signal generation, execution layers, and risk management dashboards — significantly reducing the build-your-own overhead. ## What markets are most profitable for market making? **High-volume markets with moderate uncertainty** are the sweet spot — think major political elections 2–4 weeks out, macro economic releases (Fed decisions, CPI), and major sports events during active trading periods. Very low-volume niche markets are too illiquid to fill both sides reliably. --- ## Start Market Making Smarter with PredictEngine Market making on prediction markets rewards the prepared and punishes the passive. The power users winning consistently aren't necessarily smarter — they're more systematic. They've defined their spread rules, automated their execution, capped their inventory risk, and iterated relentlessly on their data. [PredictEngine](/) is built for exactly this kind of sophisticated operation — combining real-time market data, algorithmic execution support, and analytics tools that give power users a genuine edge over manual competitors. Whether you're just formalizing your first market making strategy or scaling an existing automated book, PredictEngine has the infrastructure to match your ambition. **Explore the platform today and turn liquidity provision into a repeatable, data-driven income stream.**