Risk Analysis of Market Making on Prediction Markets: Step by Step
Market making on prediction markets means posting both buy and sell prices simultaneously, profiting from the spread — but it exposes you to inventory risk, adverse selection, and event-driven volatility that can wipe out weeks of gains in a single news cycle. Understanding these risks before you commit capital is not optional; it is the difference between running a sustainable book and blowing up your account. This guide walks you through a rigorous, step-by-step risk analysis framework built specifically for prediction market structure.
What Is Market Making in Prediction Markets?
A market maker provides liquidity by continuously quoting a bid price (the price they will buy at) and an ask price (the price they will sell at). The difference is the spread, and it is the market maker's primary source of revenue. On traditional exchanges, market makers deal in stocks or commodities. On prediction markets like Polymarket or Kalshi, the "asset" is a binary outcome: something either happens (resolves YES at $1.00) or it does not (resolves NO at $0.00).
That binary structure changes the risk profile dramatically. A stock can drift sideways for months. A prediction market contract expires at exactly 0 or 1 — and resolution can happen instantly when news breaks. This is why risk analysis for prediction market makers deserves its own framework, separate from traditional market making theory.
Step 1 — Identify Your Core Risk Categories
Before placing your first quote, you need to map the landscape of risks you are taking on. There are five primary categories:
- Inventory Risk — You hold a net long or short position when trades are imbalanced.
- Adverse Selection Risk — Informed traders hit your quotes right before prices move sharply.
- Resolution Risk — The event resolves against your position before you can flatten.
- Liquidity Risk — You cannot exit or hedge your position in a thin market.
- Operational Risk — API failures, bot errors, or platform outages freeze your book at the wrong moment.
Most beginners focus only on spreads, ignoring the other four. Professional market makers on platforms like PredictEngine weight all five when sizing positions and setting parameters.
Step 2 — Calculate Your Theoretical Edge
Your edge as a market maker comes from the spread you capture, minus the expected losses from adverse selection and inventory drift. A simple formula:
Expected PnL per round-trip = (Spread / 2) − (Adverse Selection Rate × Price Move)
For example, if you are quoting a 4-cent spread (buying at $0.48, selling at $0.52) on an event priced around 50%, and 30% of your trades are from informed traders who cause a 10-cent move against you:
- Gross per trade: $0.02
- Expected adverse selection loss: 0.30 × $0.10 = $0.03
- Net: −$0.01 per trade — you are losing money despite collecting the spread.
This math is why adverse selection rate is the single most important variable to estimate before you make a market. High-volume political markets like presidential elections often have adverse selection rates above 40%, especially in the final weeks. If you are looking at political markets specifically, the Trader Playbook: Political Prediction Markets With $10k covers how experienced traders think about edge in that environment.
Step 3 — Analyze Inventory Risk Quantitatively
Inventory risk accumulates when one side of your book gets hit more than the other. If you sell 100 YES contracts and buy only 30 back, you are net short 70 contracts. If the probability then moves from 50% to 65%, your mark-to-market loss is 70 × $0.15 = $10.50 per $1 face value in your position.
Setting Inventory Limits
Most professional market makers set hard inventory limits — maximum net exposure in contracts or dollar value. A practical framework:
| Position Size | Risk Tier | Suggested Action |
|---|---|---|
| 0–25% of limit | Low | Continue quoting normally |
| 25–60% of limit | Medium | Widen spread, reduce quote size |
| 60–85% of limit | High | Quote one side only or pause |
| 85–100% of limit | Critical | Pull quotes, actively unwind |
| >100% of limit | Breach | Emergency unwind, post-mortem |
This tiered approach is standard practice. Automating it is the next step — tools that use reinforcement learning to dynamically adjust quoting parameters are increasingly common. Automating RL Prediction Trading Explained Simply is a solid primer if you want to understand how algorithmic logic can handle these decisions in real time.
Skewing Quotes to Manage Inventory
When you accumulate net inventory on one side, the standard technique is to skew your quotes. If you are long 60 YES contracts, you shift both your bid and ask downward slightly — making it cheaper for others to sell YES back to you and more expensive to buy. This naturally attracts offsetting flow without you having to actively trade out.
Step 4 — Assess Adverse Selection Risk by Market Type
Not all prediction markets carry the same adverse selection risk. The key variable is information asymmetry: how much more do informed participants know versus the public at the moment they trade with you?
Market Type vs. Adverse Selection Risk
| Market Type | Adverse Selection Risk | Reason |
|---|---|---|
| Long-dated political (>60 days) | Medium | Information dispersed over time |
| Short-dated political (<7 days) | Very High | Insiders, polls, leaks |
| Sports events (pre-game) | Medium-High | Sharp bettors with models |
| Economic data releases | Extreme | Professional traders with fast feeds |
| Weather/climate events | Low-Medium | Mostly model-based, public data |
| Supreme Court rulings | High | Legal experts with strong priors |
Economic data release markets are particularly dangerous. A single jobs report or Fed decision can move a contract 40 cents in seconds. For a real example of how information asymmetry creates exploitable patterns — from the other side of the trade — read the Supreme Court Rulings & Arbitrage: Real Market Case Study.
Weather markets tend to have lower adverse selection because most participants are working from the same publicly available forecast models. The Weather & Climate Prediction Markets: Complete 2026 Guide explains why these markets are increasingly attractive for systematic strategies.
Step 5 — Model Resolution Risk
Resolution risk is unique to prediction markets and has no direct equivalent in equity market making. When an event resolves, every open position settles immediately at 0 or 1. If you are sitting on a 70-contract net short when a "YES" resolution comes in, you absorb the full loss with zero opportunity to exit.
Steps to Quantify Resolution Risk:
- Identify the resolution trigger — Is it a vote, a data release, a sports result?
- Estimate resolution probability — Current market price is your baseline; adjust with your own analysis.
- Calculate worst-case loss — Net inventory × (1 − current price) for longs, or Net inventory × current price for shorts.
- Determine time to resolution — The closer the event, the less time you have to rebalance.
- Set a hard close-out rule — Many market makers pull all quotes 24–72 hours before a major resolution.
- Stress test — Run a scenario where the current 40% probability event resolves YES. What is your P&L?
For time-sensitive markets, even a few hours of unmanaged inventory can be catastrophic. This is why automated position monitoring is not a luxury — it is a necessity.
Step 6 — Quantify Liquidity and Platform Risk
Thin markets create a trap: you can accumulate inventory easily, but unwinding it moves the market against you. Before you commit to making a market, check:
- Average daily volume — Below $5,000/day, your own unwinding will be market-moving.
- Order book depth — How many cents deep is the book on each side?
- Platform withdrawal mechanics — Some platforms have processing delays that affect your ability to move capital quickly.
- API reliability — If your bot drops connection, who cancels your open orders?
Cross-platform strategies can help mitigate single-platform liquidity risk. If you are making markets on Polymarket and can offset inventory on Kalshi, your effective liquidity pool is much deeper. The Cross-Platform Prediction Arbitrage via API: Profit Guide covers the mechanics of running multi-platform books with API connectivity.
For platform-specific order mechanics, the Kalshi Limit Orders: Quick Reference Guide for Traders is worth bookmarking — knowing exactly how limit orders queue and match affects your quoting strategy significantly.
Step 7 — Build Your Risk Dashboard and Ongoing Monitoring
A risk analysis completed once before you start is not enough. Market making risk is dynamic — it changes with every trade, every news event, and every shift in market consensus. Your monitoring setup should track:
- Net inventory in real time, by market
- Mark-to-market P&L updated continuously
- Time-to-resolution countdown for all active markets
- Spread captured vs. adverse selection losses (rolling 24-hour window)
- Correlation between positions — if you are long YES on two correlated political outcomes, your net risk is higher than each position in isolation
Many serious market makers use LLM-powered signal tools to scan news feeds and alert them when information breaks that could spike adverse selection risk. LLM-Powered Trade Signals: A Simple Deep Dive explains how these systems work and how they are being integrated into prediction market workflows.
Capital Allocation Rules
Never deploy 100% of your capital into market making positions. A practical structure used by experienced market makers:
- 50–60% — Active market making capital (in quotes and inventory)
- 20–30% — Reserve for inventory unwinding and emergency rebalancing
- 10–20% — Idle buffer for unexpected opportunities or margin calls
Common Market Making Mistakes and How to Avoid Them
Even with a solid framework, there are recurring errors that cause losses:
- Quoting too tight in high adverse-selection markets — Your spread is your insurance premium. Charge enough to cover expected losses.
- Ignoring correlation between positions — Being long YES on a Democrat winning both the Senate and the Presidency is not two independent bets.
- Failing to account for platform fees — A 1% fee on each side of a trade can consume your entire spread on tight markets.
- Not adjusting for market sentiment shifts — If Twitter or a major news outlet picks up a story, adverse selection risk spikes immediately.
- Manual management at scale — Trying to manage more than 3–5 markets manually in real time leads to errors. Automation is essential beyond that point.
Frequently Asked Questions
What is the biggest risk of market making on prediction markets?
Adverse selection combined with resolution risk is the most dangerous combination. Informed traders systematically hit your quotes right before a significant price move, leaving you with an inventory position that then gets locked in when the event resolves against you. Managing these two risks simultaneously is what separates profitable market makers from losing ones.
How much capital do I need to start market making on prediction markets?
There is no universal minimum, but most practitioners recommend starting with at least $2,000–$5,000 to diversify across multiple markets and maintain an adequate reserve buffer. Starting smaller means a single adverse event can wipe out your entire book before you learn the mechanics.
Can I automate market making on prediction markets?
Yes, and for serious practitioners, automation is essentially required. You need real-time inventory tracking, dynamic quote adjustment, and position monitoring that no human can sustain manually across multiple markets simultaneously. API access varies by platform — Polymarket and Kalshi both offer programmatic access with varying levels of functionality.
How do I calculate the right spread to quote?
Your spread should cover at minimum your expected adverse selection losses plus platform fees, with a margin for inventory risk. A starting formula: Spread ≥ 2 × (Adverse Selection Rate × Expected Price Move) + Platform Fee. In practice, most market makers test different spread levels empirically and adjust based on observed fill rates and P&L.
Is market making on prediction markets legal?
In the United States, legality depends on the platform and the market. Kalshi operates under CFTC regulation, making it fully legal for US participants. Polymarket restricts US users. Always verify the regulatory status of any platform in your jurisdiction before committing capital.
How does market making differ from arbitrage on prediction markets?
Market making profits from the spread between bids and asks on a single platform, taking on inventory and adverse selection risk in exchange. Arbitrage profits from price discrepancies between platforms, with theoretically lower directional risk. In practice, many sophisticated participants combine both strategies — collecting spread where possible and hedging inventory across platforms via arbitrage when discrepancies arise.
Start Making Better Risk Decisions Today
Market making on prediction markets offers a compelling edge — but only when you approach it with rigorous risk analysis at every step. From calculating your theoretical edge and modeling adverse selection to setting inventory limits and building a real-time monitoring dashboard, every element of the framework above is designed to keep your book profitable across hundreds of trades and dozens of market events.
PredictEngine brings together the tools, data, and market access you need to execute a professional market making strategy — whether you are just starting out or scaling a sophisticated multi-market book. Explore the platform, review the analytics suite, and start applying this framework to real markets today.
Run your own quant desk
Felix lets you research, backtest and run automated trading strategies, with risk controls built in.
Try Felix