A recent working paper from scholars at the London Business School and Yale University offers fresh insight into how modern prediction markets really work. By poring over years of transaction records from Polymarket—the so-called “decentralized” platform for betting on real-world events—the researchers discovered that roughly 3 percent of user accounts generate the vast majority of meaningful price adjustments.
Far from relying on broad “crowd wisdom,” the Polymarket’s accuracy and efficiency appear driven by a small, highly skilled minority.
The research study, titled “Prediction Market Accuracy: Crowd Wisdom or Informed Traders?”, examined every trade placed between 2023 and 2025. It encompasses data from 1.72 million accounts, more than 210,000 individual markets, and roughly $13.8 billion in total trading volume.
The authors—Roberto Gómez-Cram, Yunhan Guo, and Howard Kung of London Business School, together with Yale’s Theis Ingerslev Jensen—developed a rigorous method to classify traders based on how consistently their orders moved prices in the right direction and aligned with eventual outcomes.
What emerged is striking. Only about 3.14 percent of accounts qualify as “skilled winners.”
These traders reliably anticipate short-term price swings and correctly forecast final results more often than the crowd.
When liquidity providers are included, this elite group still represents under 3.5 percent of all participants yet captures more than 30 percent of the platform’s overall profits.
By contrast, roughly 67 percent of accounts are net losers who collectively absorb the platform’s entire losses.
Price discovery—the process by which new information gets translated into updated probabilities—has long been hailed as one of prediction markets’ greatest strengths.
Traditional economic theory credits the aggregated bets of many participants with producing accurate forecasts.
The new analysis flips that narrative. Instead of diffuse crowd intelligence, it is the informed actions of a tiny cohort that incorporate news, correct mispricings, and sharpen market signals.
Their trades respond faster and more accurately to breaking developments, effectively setting the benchmark that less sophisticated participants then follow.
Polymarket rose to prominence during the 2024 U.S. election cycle, drawing billions in volume as traders wagered on everything from presidential outcomes to regulatory decisions.
Its blockchain-based design promised transparency and borderless participation. Yet the paper reveals that the platform operates more like traditional financial markets than many assumed: a core group of experts profits by exploiting informational edges while the majority subsidizes their success.
For casual users, the research findings serve as a reality check. Prediction markets are not effortless wealth generators; they reward preparation, speed, and expertise.
For platform operators, the concentration of influence raises questions about liquidity design and user education.
Regulators monitoring the growth of decentralized finance may also take note: even permissionless markets exhibit familiar patterns of skill-based stratification.
The London Business School and Yale study reinforces that prediction markets remain powerful forecasting tools—but their edge comes less from collective participation and more from the quiet work of a skilled few. As these platforms mature in 2026 and attract wider audiences, recognizing this dynamic will be essential for anyone hoping to navigate them effectively.