Wisdom of the crowd: Who makes money in prediction markets?

​In recent years, prediction markets have become a popular tool for assessing the outcomes of elections, sporting events, central bank decisions, and even geopolitical conflicts. Their prices can reflect the probability of future events with considerable accuracy, but why do most participants still lose money, and who ultimately comes out ahead?

On Polymarket and Kalshi, users back their expectations with real money. If a contract trades at $0.70, the market is effectively assigning roughly a 70% probability to the corresponding event, and if it occurs, the holder receives $1.

Behind this simple mechanism lies a far more uneven distribution of outcomes.

According to a study of Polymarket trading between 2022 and 2025, around 71% of users recorded net losses, while the top 1% of traders captured roughly 95% of total profits. This creates a paradox: the market can estimate the future quite accurately even when most participants lose money.

View post on X Loading...

The tweet was deleted by the author.

But we saved everything 🙂.

Such concentration challenges the conventional idea of prediction markets as a straightforward expression of the “wisdom of the crowd.” If thousands of people collectively produce reasonably accurate probability estimates, why do so few manage to profit from the process?

Predicting the outcome is not enough

Correctly predicting an event and correctly estimating its probability are two different tasks.

Suppose a trader believes a candidate has an 80% chance of winning an election. If the corresponding contract trades at $0.60, that assessment suggests a potentially attractive opportunity. But if the price has already reached $0.90, confidence in the same candidate’s victory does not make the purchase profitable.

What matters is not only the chosen outcome, but also the contract price relative to the trader’s own probability estimate.

View post on X Loading...

The tweet was deleted by the author.

But we saved everything 🙂.

This distinction is easy to overlook because prediction markets appear simple: “yes” or “no,” one candidate winning, or a particular price being reached. Such structures encourage users to see a trade as a bet on their opinion rather than a purchase made at a specific price.

Behavioral biases reinforce the effect. A study of 287 Kalshi contracts tied to U.S. unemployment figures between July 2021 and March 2026 found that contracts priced below $0.30, representing unlikely outcomes, paid out in only around 4% of cases and were overpriced on average.

A similar pattern has long been observed in betting markets: people are willing to overpay for a small chance of a large payoff. A contract costing only a few cents may appear cheap even when its price is high relative to a reasonable estimate of the event’s probability.

Personal beliefs are another source of mistakes. Research into political prediction markets has found that participants are more likely to buy contracts tied to candidates and parties they personally support. As a result, someone can correctly predict an election winner and still make a poor trade by paying too much for that outcome.

A small group against the majority

The reasons behind this uneven distribution of profits become clearer when examining the behavior of the most successful participants.

In 2026, researchers from London Business School and Yale University analyzed transactions across 1.72 million Polymarket accounts. The authors classified around 3.14% of participants as consistently skilled.

This small group played a disproportionately large role in determining how quickly new information was incorporated into market prices.

Their advantage was not simply access to private information. These traders performed well across different markets, reacted more quickly to publicly available information, and identified discrepancies between related contracts.

Another study analyzed 588 million Polymarket trades representing $67 billion in volume. The top 1% of profitable users captured 76.5% of total profits.

The researchers also found differences in trading behavior: successful traders were more likely to provide liquidity through limit orders, while losing participants more often accepted prices already available in the market.

Moreover, professional participants do not always need to predict the future. They can identify inconsistencies between related contracts and profit from price differences regardless of the outcome.

View post on X Loading...

The tweet was deleted by the author.

But we saved everything 🙂.

In mature markets, such opportunities disappear quickly. A study of 173 NBA game markets on Polymarket analyzed more than 75 million order-book states and identified only seven executable instances of simple arbitrage during games.

The median duration of such a discrepancy was just 3.6 seconds.

A user manually assessing an event may therefore be competing not with another person reading the same news, but with professional traders and automated systems.

The market gets smarter through its losers

The classic idea of the “wisdom of the crowd” assumes that many independent estimates can collectively produce a more accurate result.

Polymarket data, however, points to a somewhat different mechanism. A large share of participants provides a diversity of opinions and a continuous flow of trades, while a small group systematically trades against the resulting errors.

Researchers describe this as the “wisdom of the informed minority.”

Skilled accounts consistently predicted both short-term price movements and the ultimate outcomes of contracts. At the same time, 67% of accounts classified as unskilled or unlucky participants generated all aggregate losses in the sample.

Each such trade simultaneously redistributes money and helps correct the price. If a contract is overpriced, better-prepared participants sell it or take the opposite side. Their actions gradually reduce the gap between the market price and the information available.

View post on X Loading...

The tweet was deleted by the author.

But we saved everything 🙂.

This creates an unusual system: a market does not need most of its participants to be good at predicting the future. It only requires a relatively small group capable of spotting other participants’ mistakes and financially motivated to trade against them.

For an ordinary user, the problem is therefore not necessarily a lack of understanding of politics, sports, or economics. They may correctly identify the eventual winner but pay too much for the contract, react to information too slowly, or choose a less efficient way to execute the trade.

This is the central contradiction of prediction markets. The system as a whole can remain accurate while most participants lose money because their mistakes create opportunities for a small group of more efficient traders.

In that sense, the “wisdom of the crowd” is not the sum of equally good predictions. Instead, it emerges because a small group of participants continuously identifies and corrects the mistakes made by the majority.