Market Making on Prediction Markets: Real Case Studies
Market making on prediction markets is one of the most consistent — yet underappreciated — ways to generate returns in the space. By continuously quoting both buy and sell prices on binary outcomes, skilled market makers capture the bid-ask spread while managing directional exposure. This guide breaks down real case studies, showing exactly how traders have profited from liquidity provision on platforms like Polymarket.
Whether you're a quant-curious newcomer or an experienced trader looking to diversify into prediction markets, understanding how market making actually works in practice — not theory — is the fastest way to sharpen your edge. Let's dig into the mechanics, the numbers, and the lessons learned from real trading activity.
What Is Market Making on Prediction Markets?
Market making is the act of simultaneously posting bids and offers on a market, profiting from the difference between the two prices (the bid-ask spread). On traditional exchanges, this is done by firms like Citadel or Virtu. On prediction markets, individuals and bots can step into this role.
In a binary prediction market, outcomes are priced between $0 and $1 (or 0¢ and 100¢). A market maker might post:
- Bid: Buy "Yes" at 42¢
- Ask: Sell "Yes" at 46¢
If both sides fill, the maker earns 4¢ per share — regardless of the actual outcome, assuming balanced flow. The catch? If the underlying probability shifts, you can end up holding inventory on the wrong side.
For a deeper dive into the mechanics before reading the case studies, check out this comprehensive market making guide for power users.
Case Study #1: The Federal Reserve Rate Decision Market (March 2024)
Setup
In early March 2024, Polymarket listed a market asking: "Will the Fed cut rates at the March 2024 FOMC meeting?" This was a high-attention market with significant volume — over $4.2 million traded before the resolution date.
A trader using a semi-automated quoting bot set up the following:
- Initial spread: Bid at 8¢, Ask at 13¢ (5¢ spread)
- Position size per order: $500 notional
- Target daily volume: 40–60 fills
What Happened
As Fed commentary shifted hawkish through the month, the true probability dropped from roughly 15% to under 5%. The market maker's bot was programmed with a skew adjustment: anytime inventory on the "Yes" side exceeded $1,200, it would widen the ask and narrow the bid to reduce further Yes accumulation.
This prevented a blow-up. By mid-March, the trader had:
- Completed 312 round-trip fills
- Gross spread income: $1,840
- Inventory loss from directional drift: -$290
- Net profit: ~$1,550 over 18 trading days
For those interested in how Fed markets behave specifically, see Fed Rate Decision Markets: Best Practices with PredictEngine for additional context on positioning around FOMC events.
Key Lesson
Dynamic spread widening during inventory buildup is essential. Static quoting in fast-moving markets bleeds edge rapidly.
Case Study #2: Sports Event Market Making During NBA Playoffs
Setup
The 2024 NBA Playoffs generated a wave of game-by-game prediction markets. One trader focused specifically on series outcome markets (e.g., "Will the Celtics win the series?") rather than game-by-game results, reasoning that series markets have slower information decay.
Their approach:
- Identify markets with over $500K in expected volume
- Quote a 3–5¢ spread around the implied probability
- Rebalance inventory every 4 hours using closing line value (CLV) as a benchmark
- Hard stop: exit if net directional exposure exceeds $3,000
Results Over 6 Weeks
| Metric | Value |
|---|---|
| Total markets made | 14 |
| Total volume facilitated | $387,000 |
| Gross spread captured | $9,200 |
| Hedging/inventory losses | -$2,100 |
| Platform fees paid | -$780 |
| Net profit | $6,320 |
| Return on capital deployed | ~8.4% (6 weeks) |
The trader's biggest win was the Knicks-Pacers series market, where chaotic public sentiment created wide, inefficient spreads for 3 straight days. Their bot captured over $1,800 in that single market.
Want to understand how crypto and sports intersect in prediction market flow? The trader playbook on Bitcoin predictions during NBA Playoffs is a fascinating companion read.
Case Study #3: Ethereum Price Prediction Markets
Setup
Crypto price prediction markets ("Will ETH close above $3,500 on June 30?") are volatile and informationally rich — a challenging but profitable environment for market makers who manage risk well.
One team of two traders deployed a simple Greeks-aware inventory model:
- Quote 4¢ spreads on weekly ETH close markets
- Hedge net "Yes" exposure by shorting ETH perpetual futures (delta hedging)
- Size each order at $250 max, scaling down as IV rose
Results
Over Q1 2024, they tracked performance across 11 separate ETH price markets:
- Best week: $2,200 net profit (ETH was range-bound, spreads were sticky)
- Worst week: -$640 (ETH spiked 12% after a surprise ETF filing — hedges lagged)
- Average weekly net: +$880
- Sharpe ratio estimate: ~1.8 annualized
The delta hedging strategy cut directional losses by an estimated 60% compared to unhedged runs. The main risk was slippage on the hedge leg — when ETH moved fast, the perp fills came in 0.3–0.8% worse than expected.
For more on managing these dynamics, see the Ethereum Price Predictions & Limit Orders: Real Case Study and this detailed breakdown of algorithmic slippage in prediction markets.
How to Start Market Making on Prediction Markets: Step-by-Step
Here's a practical framework based on what actually worked in the case studies above:
- Choose a high-volume market niche — Fed decisions, crypto price markets, and major sports series all generate the volume needed for spread capture to be meaningful.
- Set your initial spread width — Start with 5–8¢ on markets where true probability is between 20–80%. Avoid sub-10% or over-90% markets as a beginner; the risk of adverse selection spikes.
- Define your inventory limits — Decide your maximum acceptable directional exposure before you start. For most retail market makers, $1,000–$2,500 per market is reasonable.
- Build or rent a quoting bot — Manual quoting is nearly impossible at scale. Use PredictEngine or a compatible API framework to automate order placement.
- Implement a skew rule — As inventory builds on one side, shift your quotes to discourage further accumulation. A simple rule: for every $500 of net inventory, widen the disadvantaged side by 1¢.
- Hedge when necessary — For crypto markets, delta hedge with perpetuals. For political or sports markets, consider correlated positions in adjacent markets.
- Track CLV (Closing Line Value) — If you're consistently posting prices that get beaten by the closing line, you're being adversely selected. Tighten your model or pull back.
- Review and iterate weekly — Look at fill rates, inventory profiles, and net PnL per market. Drop underperformers. Double down on efficient markets.
Common Pitfalls and How Real Traders Avoided Them
Adverse Selection From Sharp Flow
The single biggest risk for prediction market makers is getting picked off by informed traders who know the answer before the market does. In the Fed case study, the bot detected unusual fill skew (too many "Yes" buyers in a row) and paused quoting for 15 minutes — a simple but effective circuit breaker.
Fee Drag on Tight Markets
Platform fees on Polymarket typically run 2% of winnings. On a 4¢ spread in a 50/50 market, fees can eat 20–40% of gross spread income if you're not careful. The NBA trader above paid $780 in fees on $9,200 gross — an 8.5% drag, which was manageable only because volume was high.
Liquidity Illusions
Some markets look liquid but have concentrated positions from a few whales. If a whale exits, the spread blows wide and your inventory is suddenly very wrong. Cross-reference order book depth before committing capital.
Comparing Market Making Approaches: Manual vs. Automated
| Approach | Setup Time | Scalability | Spread Consistency | Risk of Error |
|---|---|---|---|---|
| Manual quoting | Low | Very low | Poor | High (fatigue) |
| Semi-automated (alerts + manual) | Medium | Medium | Moderate | Medium |
| Fully automated bot | High | Very high | Excellent | Low (if coded well) |
| Platform tool (e.g., PredictEngine) | Low | High | Good | Low |
For most traders, the semi-automated to fully automated path is optimal. PredictEngine offers tools that bridge this gap — letting you set quoting rules, inventory limits, and spread logic without writing code from scratch.
What Returns Should You Realistically Expect?
Based on the case studies and community data across experienced prediction market makers:
- Beginner market makers (manual or basic automation): 2–5% monthly ROI on deployed capital, with high variance
- Intermediate makers (automated quoting, basic inventory management): 5–10% monthly, more consistent
- Advanced makers (delta hedging, dynamic spread, multi-market): 10–20%+ monthly, but requires significant infrastructure
These numbers assume active management, reasonable capital ($5,000–$25,000), and focus on liquid markets. Scaling beyond $50,000 in a single market becomes difficult without moving the market yourself.
For context on how portfolio-level risk management interacts with these returns, see this real-world portfolio hedging case study.
Frequently Asked Questions
What is market making in prediction markets?
Market making in prediction markets involves continuously quoting both buy and sell prices on binary outcomes, profiting from the bid-ask spread. Unlike directional trading, market makers don't need to predict the correct outcome — they profit from facilitating trades between other participants. The key risk is holding inventory on the wrong side when probabilities shift rapidly.
How much capital do I need to start market making on prediction markets?
Most practitioners recommend starting with at least $2,000–$5,000 in active capital to generate meaningful spread income while maintaining adequate inventory buffers. Below $1,000, fee drag and minimum order sizes make consistent profitability difficult. The case studies above used $5,000–$15,000 in deployed capital.
What's the biggest risk in prediction market making?
Adverse selection is the primary risk — being repeatedly filled by traders who have better information than you. The second major risk is inventory blowup, where you accumulate a large directional position right before the market moves against you. Both can be mitigated with inventory limits, circuit breakers, and dynamic spread adjustments.
Can I automate prediction market making?
Yes, and for most serious market makers, automation is essential. Tools like PredictEngine allow traders to set automated quoting rules, manage inventory, and track performance across multiple markets simultaneously. The traders in the NBA Playoffs case study used a semi-automated bot that handled fill tracking and rebalancing with minimal manual input.
How do platform fees affect market making profitability?
Platform fees directly reduce gross spread income and can be significant on tight markets. On Polymarket, fees of 2% of winnings are standard. If your average spread is only 4¢, fees may consume 25–40% of gross income depending on market dynamics. Focus on markets where you can maintain 5¢+ spreads to keep fees below 20% of gross.
Is prediction market making legal?
In most jurisdictions, trading on regulated prediction markets is legal for individuals. Polymarket operates under CFTC oversight in the US for certain products, and access rules vary by country. Always verify your local regulations before deploying capital. Market making itself is a recognized and legal trading strategy, used by professional firms across all asset classes.
Start Market Making With the Right Tools
Market making on prediction markets is one of the few strategies where disciplined execution consistently beats pure prediction skill. The case studies above prove it: with the right inventory rules, automation, and risk management, traders are pulling in consistent monthly returns of 5–15% — without needing to call every political or sports outcome correctly.
The key is getting your infrastructure right from day one. PredictEngine gives you the quoting tools, analytics, and automation framework that make the strategies in this article achievable without a full engineering team behind you. Whether you're starting with $3,000 or scaling past $50,000, the platform is built for exactly this kind of systematic, spread-based trading. Start your free trial today and see how far a disciplined market making strategy can take you.
Run your own quant desk
Felix lets you research, backtest and run automated trading strategies, with risk controls built in.
Try Felix