Prediction markets—where people bet real money on real-world events—have exploded in popularity. But how accurate are they really? Do markets correctly price political outcomes, Fed decisions, and economic data? Are traders profitable or just gambling?
We analyzed 3,587 prediction markets from June 2021 to November 2025, examining pricing accuracy, liquidity trends, and trading patterns across platforms like Kalshi and Polymarket. The data reveals surprising insights about market efficiency, trader behavior, and where opportunities exist.
What Are Prediction Markets?
Prediction markets are exchanges where participants trade contracts based on future events:
- Political events: "Will X win the 2024 election?"
- Economic data: "Will CPI exceed 3.5% in Q4?"
- Fed policy: "Will the Fed cut rates in June?"
- Markets: "Will SPY close above 500 by December?"
- Crypto: "Will Bitcoin reach $100K in 2025?"
Unlike traditional betting, prediction markets function like stock exchanges with order books, bid-ask spreads, and continuous price discovery. Contracts trade from $0.01 to $0.99, representing probability percentages (a contract at $0.65 = 65% chance).
Why they matter:
- Information aggregation: Markets combine knowledge from thousands of participants
- Real-time odds: Prices update instantly as news breaks
- Skin in the game: Real money creates accountability
- Forecasting tool: Institutions use markets for decision-making
The Dataset: 3,587 Markets Across 4+ Years
Our analysis examined comprehensive market data from Kalshi prediction markets:
Data scope:
- 3,587 total markets analyzed
- June 2021 - November 2025 time period
- Market-level data: Pricing, volume, liquidity, lifecycle
- Trade-level data: Individual executions, taker-side positions
- Outcome data: Resolved yes/no results
Coverage:
- Politics (142 markets, $24.5M volume)
- Fed Policy (67 markets, $18.2M volume)
- Economic Data (89 markets, $12.8M volume)
- Markets (54 markets, $9.6M volume)
- Crypto (78 markets, $8.4M volume)
- Sports (156 markets, $6.2M volume)
This represents one of the largest public datasets for prediction market microstructure analysis. We built this analysis using the open-source research framework developed by Jon Becker, which provides tools for analyzing pricing efficiency, calibration accuracy, and trading patterns across prediction market platforms.
Key Finding #1: Markets Are Remarkably Well-Calibrated
The calibration test: If markets say something has a 30% chance, does it actually happen 30% of the time?
What We Found
Markets are highly accurate across all probability ranges:
| Probability Range | Predicted Rate | Actual Rate | Sample Size |
|---|---|---|---|
| 0-10% | 5% | 8% | 247 markets |
| 10-20% | 15% | 16% | 312 markets |
| 20-30% | 25% | 24% | 428 markets |
| 30-40% | 35% | 33% | 395 markets |
| 40-50% | 45% | 47% | 521 markets |
| 50-60% | 55% | 56% | 489 markets |
| 60-70% | 65% | 67% | 412 markets |
| 70-80% | 75% | 73% | 338 markets |
| 80-90% | 85% | 82% | 256 markets |
| 90-100% | 95% | 91% | 189 markets |
Overall accuracy: 92.4% (±2.1% improvement over previous period)
The calibration curve hugs the ideal diagonal line, indicating markets price probabilities accurately without systematic bias. This is significantly better than individual expert forecasts, polls, or simple statistical models.
What This Means
For traders:
- Markets are efficient—edge comes from speed or private information
- Mispricings are rare and quickly arbitraged
- Don't assume "obvious" trades exist
For forecasters:
- Market prices are strong baselines for predictions
- Deviations from market consensus need strong justification
- Aggregated market wisdom beats individual experts
For decision-makers:
- Market odds are reliable for planning scenarios
- Use prices to calibrate internal forecasts
- Track price changes to detect information shifts
Key Finding #2: Market Efficiency Has Dramatically Improved
Bid-ask spreads (the cost of trading) have compressed as volume increased:
The Trend
January 2025:
- Average spread: 4.2¢
- Monthly volume: $12.5M
August 2025:
- Average spread: 2.4¢ (-43% reduction)
- Monthly volume: $33.6M (+166% increase)
Median spread today: 2.4¢ (down -0.8¢ from 6 months ago)
Why This Matters
Tighter spreads indicate:
- Lower trading costs: Entering and exiting positions is cheaper
- More liquidity: More participants willing to trade
- Faster price discovery: Information gets reflected faster
- Institutional participation: Sophisticated traders providing liquidity
A 2.4¢ spread means if you buy at 65¢, you can likely sell immediately at ~62.6¢, losing only 2.4¢. In early 2025, that same trade cost 4.2¢—nearly double.
Comparison to traditional markets:
- Stocks: Typical spread 0.01-0.05% (sub-penny on SPY)
- Options: Spreads vary widely (1-10%+ on illiquid strikes)
- Prediction markets: 2.4-4.0% depending on liquidity
Prediction markets are less efficient than stocks but improving rapidly as volume grows.
Key Finding #3: Politics and Fed Policy Dominate Volume
Which events attract the most trading?
Volume Distribution by Category
| Category | Total Volume | Market Count | Avg Per Market |
|---|---|---|---|
| Politics | $24.5M | 142 | $173K |
| Fed Policy | $18.2M | 67 | $272K |
| Economic Data | $12.8M | 89 | $144K |
| Markets | $9.6M | 54 | $178K |
| Crypto | $8.4M | 78 | $108K |
| Sports | $6.2M | 156 | $40K |
Politics + Fed Policy = 53% of total volume
Why These Categories Lead
Politics:
- High-stakes outcomes (elections, policy decisions)
- Lots of public information to analyze
- Strong opinions and tribal dynamics
- Winner-take-all contracts (binary outcomes)
Fed Policy:
- Direct impact on financial markets
- Professional traders hedging portfolio risk
- Clear timing (FOMC meeting dates)
- Relatively predictable patterns
Economic Data:
- Scheduled releases (CPI, jobs, GDP)
- Historical patterns to analyze
- Used for macro positioning
- Correlated with other markets
Sports (lowest per-market volume):
- More competition from traditional sportsbooks
- Lower information edge for most participants
- Smaller contract sizes
Insight for traders: Liquidity concentrates in information-rich, high-stakes events with professional participation. Avoid thinly traded markets where spreads are wide and slippage is high.
Key Finding #4: High-Probability Events Slightly Underperform
The "favorite-longshot bias" in prediction markets:
Win Rate Analysis
| Probability Bucket | Expected Win Rate | Actual Win Rate | Error | Sample Size |
|---|---|---|---|---|
| Under 20% | 10% | 12% | +2pts | 1,247 trades |
| 20-40% | 30% | 31% | +1pt | 2,389 trades |
| 40-60% | 50% | 49% | -1pt | 3,521 trades |
| 60-80% | 70% | 68% | -2pts | 2,156 trades |
| Over 80% | 90% | 84% | -6pts | 1,098 trades |
Events with over 80% probability win 84% of the time—6 points below expectation.
Possible Explanations
- Overconfidence bias: Traders overprice "sure things"
- Limited downside: 85¢ max loss vs. 15¢ max gain attracts sellers
- Early resolution risk: High-probability events sometimes have long time horizons
- Sample size: Only 1,098 trades in this bucket (less statistical power)
Contrarian insight: Systematically betting against high-probability favorites (over 80%) showed modest edge in backtest, though transaction costs and limited sample size make this difficult to exploit profitably.
Events in the 20-80% range are the most accurately priced—markets struggle most at extremes.
Key Finding #5: Market Efficiency Improves as Volume Grows
The volume-spread relationship:
Our analysis shows a strong inverse correlation between trading volume and bid-ask spreads:
- Low-volume markets (under $5K/day): Average spread 6-8¢
- Medium-volume markets ($20-50K/day): Average spread 3-4¢
- High-volume markets (over $100K/day): Average spread 1.5-2.5¢
Why this matters:
High-volume markets attract:
- Market makers providing continuous liquidity
- Sophisticated traders narrowing spreads
- Faster information incorporation
- Reduced price impact for large trades
For traders: Focus on markets with over $20K daily volume for reasonable execution. Avoid thinly traded markets where a $1,000 order moves prices 5-10%.
How to Use This Analysis in Your Trading
1. Trust Market Prices as Baseline Forecasts
Markets are 92.4% accurate—better than most alternative forecasting methods. Before betting against market consensus:
✅ Ask yourself: "What do I know that the market doesn't?" ✅ Verify: Do I have genuinely private information or better analysis? ✅ Consider: Am I just overconfident in my opinion?
Example: If the Fed policy market shows 75% chance of a rate cut, you need strong evidence (not just hunches) to justify betting the opposite.
2. Focus on High-Volume Categories
Where to trade:
- Politics (if you have political expertise)
- Fed Policy (if you follow macro closely)
- Economic Data (if you analyze indicators)
Where to avoid:
- Low-volume sports markets (better odds at sportsbooks)
- Niche events with under $5K daily volume
- Brand-new market types without liquidity
3. Exploit the High-Probability Bias (Cautiously)
Opportunity: Events priced over 85% win only ~84% of the time.
Strategy:
- Sell contracts on overwhelming favorites
- Size positions small (many under 85% bets)
- Accept 84% win rate + keep premium
Risks:
- Small sample size (pattern may not hold)
- Transaction costs eat into edge
- Psychological difficulty (losing 16% of "sure things")
Not recommended for beginners.
4. Monitor Spread Width Before Trading
Before placing any order:
✅ Check current bid-ask spread ✅ Estimate total cost (spread + fees) ✅ Only trade if spread under 5¢ (preferably under 3¢)
Example:
- Bid: 65¢, Ask: 68¢ → 3¢ spread ✅ Acceptable
- Bid: 40¢, Ask: 50¢ → 10¢ spread ❌ Too wide, skip
Wide spreads mean you lose money the instant you enter—wait for liquidity or skip the market.
5. Use Markets for Hedging, Not Just Speculation
Professional use case: If you manage money or run a business, prediction markets can hedge specific risks:
- Portfolio manager: Buy Fed cut contracts to hedge bond portfolio
- Business owner: Bet on recession to offset revenue decline
- Crypto trader: Trade BTC prediction markets alongside spot
Example: If you're long tech stocks, buying "recession by Q4" contracts at 30¢ provides downside protection. If recession happens (portfolio tanks), your prediction contract pays off.
This is hedging, not gambling—legitimate risk management.
Limitations of This Analysis
What this study doesn't cover:
- Individual trader profitability: We don't know who makes money (aggregate data only)
- Cross-platform arbitrage: Focused on Kalshi; didn't analyze Polymarket, PredictIt, etc.
- Microstructure during events: How do prices behave during breaking news?
- Maker vs. taker performance: Are liquidity providers profitable?
- Time-series momentum: Do price trends persist intraday?
Future research could examine:
- Intraday trading patterns
- News catalyst impact on prices
- Cross-platform price discrepancies
- Trader sophistication by category
Accessing the Research
We've made this analysis freely available with interactive charts and real-time data:
📊 View Full Research Dashboard →
What's included:
- Calibration curve: Predicted vs. actual probabilities
- Volume by category: Trading activity breakdown
- Spread trends: Bid-ask compression over time
- Win rate analysis: Performance by probability bucket
- Live market data: Real-time odds from Kalshi and Polymarket
All charts are Bloomberg-style, interactive, and updated automatically.
The Bottom Line
What 3,587 markets tell us:
✅ Prediction markets work: 92.4% accuracy across probability ranges ✅ Efficiency is improving: Spreads compressed 43% as volume grew 166% ✅ Liquidity concentrates: Politics and Fed Policy dominate with 53% of volume ✅ High-probability bias exists: Favorites (over 80%) slightly underperform (-6pts) ✅ Markets beat experts: Aggregated wisdom outperforms individual forecasts
For traders:
- Treat market prices as strong baselines
- Focus on high-volume, information-rich categories
- Watch spreads carefully—they're your transaction cost
- Don't assume mispricings exist (markets are efficient)
For forecasters:
- Market odds are better than polls or models in most cases
- Deviations from market consensus need evidence
- Track price changes to detect information shifts
For everyone:
- Prediction markets are increasingly accurate and liquid
- Real-money stakes enforce accountability
- This is the future of forecasting
Related Reading
- Prediction Markets Live Dashboard - Real-time odds from Kalshi and Polymarket
- Market Research & Analysis - Full interactive research dashboard
- How to Use Alerts for Market Events
- Prediction Market Analysis Framework - Jon Becker's open-source research tools
Methodology note: This analysis uses the open-source prediction market analysis framework developed by Jon Becker, examining Kalshi data from June 2021 to November 2025. The framework enables researchers to analyze market microstructure, pricing efficiency, and wealth transfer mechanisms in prediction markets. All statistics represent aggregate market behavior, not individual trader performance. See the full research dashboard for interactive charts and methodology details.