The appeal — and the hidden assumption — of crowd pricing
Prediction markets are built on a simple promise. Instead of asking one expert to guess an outcome, let thousands of people trade contracts on it, and let the price settle wherever supply and demand find equilibrium. The theory traces back to the old idea that a diverse crowd, each holding a sliver of private information, can collectively outperform any single forecaster — provided the errors of individual participants are scattered rather than pointing the same direction. That is the textbook "wisdom of crowds" mechanism, and prediction-market advocates often add a second layer: because trading costs real money, people with weak opinions are supposed to bet small, and people with strong information are supposed to bet big. If that were reliably true, weighting the crowd’s aggregate price by trade size would make the forecast better, not worse.
The catch is that this second layer is an assumption, not a law of markets. Money at stake tells you how confident someone is. It does not automatically tell you whether that confidence is earned.
What the Berkeley researchers actually measured
A team from UC Berkeley set out to test that assumption directly, rather than take it on faith. They examined the internal trading data of Kalshi and Polymarket — the two dominant U.S.-linked prediction-market platforms — across 5,456 markets, grouped into three very different categories: fast-moving 15-minute crypto price bets, "mention markets" wagering on what a company might say during an earnings call, and forecasts on NBA game outcomes.
For each trader, the researchers calculated something they called "edge": how profitable that trader’s bets were on average, independent of how large the bets were. Then they sorted traders from smallest to largest by wager size and compared edge across that spectrum. The logic is straightforward — if a prediction market’s aggregate price is supposed to reflect wisdom, the people whose money moves that price the most (the largest traders, often nicknamed "whales") should be at least as accurate as everyone else. If they are actually worse, the market is effectively letting its least reliable voices shout the loudest.
That is what the data showed. In the 15-minute crypto markets, the smallest traders had the strongest edge and the largest traders the weakest. In the mention markets, whales’ edge turned negative — meaning a strategy of simply copying their trades would have lost money on average. The NBA forecasts followed the same pattern: the biggest bettors underperformed the smaller ones. Because those large trades are precisely the ones that move the price the most, the aggregate forecast ends up shaped disproportionately by the participants with the least demonstrated accuracy.
The researchers pushed one step further and asked why large traders behave this way. They found no meaningful statistical link between how strongly a trader expressed conviction (via sentiment or size) and how informed that conviction actually was. In their own words, the most prominent voices in these markets "function primarily as sources of communicative noise" rather than as carriers of superior information.
The mechanism: conviction crowds out accuracy
It helps to see the failure as a chain of small, individually plausible steps that compound into a distortion.
flowchart TD A[Large wager placed] --> B[Larger price impact per trade] B --> C[Whale opinion overweighted in aggregate price] C --> D[Little correlation between size and accuracy] D --> E[Aggregate price skewed toward noisy conviction]
Each link in that chain is mundane on its own: bigger orders naturally move prices more in any order-book market, and a market has no built-in way to check whether the trader behind a large order is well-informed or simply certain. The problem only becomes visible when you compare edge across trade sizes, which is exactly what the study did — and what most casual readings of prediction-market prices never bother to check.
Not every market fails the same way
This does not mean prediction markets are broken as a category, and the study itself does not claim that whales are wrong everywhere or that small traders are always right. What it suggests is that the conditions under which a market aggregates information well can be undermined by the same features that make it a market at all — order size, liquidity, and the incentive to trade on conviction rather than evidence.
| Condition | Supports reliable price discovery | Can distort the signal |
|---|---|---|
| Trader base | Diverse participants with varied, independent information | A small number of large, ideologically driven traders |
| Resolution criteria | Clear, objective, verifiable outcome | Ambiguous or subjective resolution, inviting speculation |
| Liquidity | Deep enough that no single order swings the price | Thin order books where one big trade sets the odds |
| Size–accuracy relationship | Larger bets correlate with demonstrated edge | Larger bets reflect conviction or capital, not accuracy |
Liquidity and platform structure matter here too, if only as context. Kalshi and Polymarket differ in design — one centralized and standardized, the other more decentralized and varied in market types — and both report that thin, low-volume markets tend to produce noisier pricing than deep ones. That is a structural observation about when markets are more or less trustworthy, not proof of who is right within them, and it should not be read as independent confirmation of the Berkeley finding itself.
The broader lesson for investors
The reason this matters beyond prediction-market enthusiasts is that the underlying mistake is not exotic. It is the same one investors make when they read a large trade, a loud public bet, or a confidently worded thesis as evidence of insight. Size and volume are visible; information quality is not. A trader — or an investor, or a pundit — can hold a large position because they know something others don’t, or because they are hedging, signaling, defending an ideology, or simply have more capital to lose. Nothing about the size of the bet distinguishes between those motives.
The Berkeley data cannot tell us how far this pattern extends beyond the specific markets studied, and it does not license the conclusion that big investors are generally less informed than small ones in every market. Traditional financial markets have different participants, different incentives, and different feedback loops, and the same size-based effect has not been demonstrated there in the same form. What the finding does offer is a sharper version of an old caution: before treating any price — a prediction-market forecast, a heavily traded stock, or a widely repeated conviction — as a signal of collective wisdom, ask what is actually driving its size. Diversity of opinion, clear stakes, and demonstrated accuracy build a trustworthy aggregate. Volume and confidence alone do not.
