How 'Superforecasters' and Prediction Markets Are Beating the Experts

Human beings are obsessed with predicting the future. When a crisis looms, an election approaches, or the stock market wavers, we instinctively turn to credentialed experts, television pundits, and specialized analysts to tell us what will happen next. But a quiet revolution in data science and behavioral economics has proven that the traditional experts are usually wrong—and that a decentralized crowd of amateurs can consistently beat them.[3][7]

The mechanism driving this shift is the prediction market. Instead of asking people what they think will happen in a poll, prediction markets ask participants to put money or reputation on the line. Platforms like Polymarket, Kalshi, and Metaculus operate like stock exchanges for real-world events. If shares for a specific outcome are trading at $0.60, the market is estimating a 60% probability that the event will occur.[1][4]

This financial incentive fundamentally changes human behavior. In surveys, people posture; on social media, they signal their political identity. But in a prediction market, accuracy is rewarded, noise is punished, and wishful thinking becomes expensive. This dynamic strips away ideology and forces participants to seek the objective truth.[1]

The sheer scale of this new forecasting infrastructure is staggering. By late 2025, trading volume on major prediction platforms surpassed $3 billion in a single quarter. Corporate leaders and policymakers are increasingly abandoning backward-looking dashboards and slow-moving polls in favor of these real-time intelligence networks, which react to new information instantly.[1][7]

The scientific rigor behind this movement traces back to the Good Judgment Project, an initiative funded by the U.S. intelligence community in 2011. Researcher Philip Tetlock hosted massive forecasting tournaments to see who could best predict complex geopolitical events. The goal was to find out if the "wisdom of the crowd" could outperform the CIA.[2][5]

The tournaments revealed a unique class of people dubbed "superforecasters." They were not necessarily subject-matter experts, nor did they have access to classified information. They were ordinary people—software engineers, pharmacists, retirees—who possessed a specific set of cognitive habits that allowed them to see the future more clearly than anyone else.[3][5]

The data from these tournaments was undeniable. Superforecasters achieved Brier scores—a strict mathematical metric of prediction accuracy—that were 20% to 30% lower than the median forecaster. In historical studies, prediction markets have accurately aligned with election outcomes 74% of the time, routinely outperforming traditional polling and punditry.[2][4]

So, what makes a superforecaster? Tetlock found that they are "foxes" rather than "hedgehogs." A hedgehog views the world through one big, unifying ideological lens, squeezing complex problems into a preferred narrative. A fox is pragmatic, gathering diverse pieces of information from multiple sources and avoiding loyalty to any single idea.[3][7]

Superforecasters also practice "Bayesian updating." When faced with a question, they start with a baseline probability—the "outside view"—and incrementally adjust it up or down as new evidence arrives. They treat their beliefs as hypotheses to be tested, rather than treasures to be guarded.[3][5]

The single strongest predictor of forecasting success is a trait known as "perpetual beta." This is a relentless commitment to self-improvement and belief updating. Researchers found that this willingness to admit mistakes and recalibrate is roughly three times as powerful a predictor of accuracy as raw intelligence.[3][7]

However, the system is not flawless. In scenarios where highly specialized insider knowledge is required—such as a pending FDA drug approval or a closed-door corporate merger—a single expert with legal access to that data will still out-predict a decentralized crowd. Markets can only aggregate the information that is publicly available.[4][5]

Recent academic studies have also identified structural biases within these platforms. For example, political prediction markets often exhibit persistent underconfidence, where prices chronically compress toward 50% due to the emotional volatility of the participants. Furthermore, markets require high liquidity; without enough active traders, the wisdom of the crowd breaks down.[6]

As prediction markets expand, they also face moral scrutiny. While betting on economic indicators or sports is widely accepted, platforms have hosted contracts on wars, assassinations, and human tragedies. This raises ethical questions about the societal impact of gamifying global crises and the potential for market manipulation.[6][7]

Despite these controversies, the shift toward quantified, accountable forecasting is permanent. The era of the unaccountable pundit is fading. By adopting the habits of superforecasters—breaking down complex problems, seeking diverse viewpoints, and embracing uncertainty—anyone can improve their ability to navigate an unpredictable world.[1][3][7]