A market price aggregates many independent views into one number. When the conditions hold, that number is hard to beat. When they break, the crowd can be confidently wrong.
A prediction market is a machine for averaging many independent judgements into one price. When the crowd is diverse and independent, that average is hard to beat. When it herds, it is not.
In Galton's 1907 study, the average of 787 guesses at an ox's weight, 1,197 pounds, fell within one pound of the true 1,198 pounds, per his note Vox Populi in Nature.
A wise crowd can still be wrong. A price near one cent or ninety nine cents is a probability, not a certainty, and whether markets beat polls is genuinely debated.
28 June 2026, reviewed by Fredrik Filipsson. This is general information, not advice, and trading carries a real risk of loss.
The wisdom of crowds is the finding that the combined judgement of many independent people can beat most individuals, and sometimes the experts. A prediction market is a machine for doing exactly that. It pools the views of many participants, each backing an opinion with money, into a single price that reads as an implied probability. The signal is strongest when several conditions hold: diversity of views, independence so people are not copying one another, decentralization that lets local knowledge in, and a mechanism to aggregate it all, which a market supplies through price and sharpens with money at stake. When independence collapses into herding, or a market is thin, the crowd can be badly wrong. The price is a serious benchmark, never a guarantee.
The writer James Surowiecki, drawing the research together in his 2004 book The Wisdom of Crowds, set out four conditions a crowd needs to be accurate. A prediction market satisfies the fourth directly, through price, and adds a fifth ingredient of its own, honest financial incentives, that a guessing jar does not have.
The crowd is smart when people bring different pieces of information and different angles, even eccentric ones. A room of identical views adds nothing. Variety is what lets errors point in different directions so they can cancel out.
People must judge for themselves rather than copy each other. When everyone follows the same loud voice, independence collapses and the crowd starts to herd, which is when collective judgement turns fragile.
People can specialise and draw on local knowledge, the detail that only someone close to the question would notice. A market lets that scattered, specific knowledge find its way into the price through trades.
Scattered opinions are useless until something combines them. A market does this through price. Each trade nudges the number, so the price becomes a running, weighted summary of what the crowd currently believes, updated the moment news arrives.
In Galton's fair, individual guesses at the ox's weight were all over the place, yet their average landed within a pound of the truth. Drag the slider to add more independent guesses to a simulated crowd and watch the running average converge toward the true value. This is an illustration of the principle, not a prediction about any real market.
Illustrative simulation. Real crowds only converge when guesses are diverse and independent. A herding crowd does not.
The idea has a long paper trail. The table gathers the milestones people cite most, with their sources and dates, so you can see where the claim comes from rather than taking it on faith. Note the last row: the evidence that markets beat polls is real but contested, and honest sources disagree about the size of the edge.
| Source and date | What it measured | The headline result |
|---|---|---|
| Galton, Vox Populi, Nature, 1907 | 787 guesses at the dressed weight of an ox at a Plymouth fair | The average guess, 1,197 pounds, fell within one pound of the true 1,198 pounds |
| Surowiecki, The Wisdom of Crowds, 2004 | A survey of when group judgement is accurate | Sets out four conditions: diversity, independence, decentralization, and aggregation |
| Iowa Electronic Markets accuracy research (Berg, Forsythe, Nelson and Rietz; long run analyses) | Election eve market prices versus final vote shares since 1988 | Average absolute error about 1.34 points for US presidential markets; the market was closer than the polls about 74 percent of the time |
| Counterpoint, ongoing academic debate | Fresh analyses of markets versus polls across periods | Some studies find the edge small or sensitive to the sample, so the advantage is real but contested |
Method and sourcing, as of 28 June 2026. Figures are quoted from the original Galton note and from published analyses of the Iowa Electronic Markets. The IEM error figures are averages across many markets and are not a guarantee about any single future election. We present the counterpoint row because the comparison of markets to polls remains an active research question, and treating it as settled would overstate the case.
Figure: average absolute election eve error across Iowa Electronic Markets since 1988, per published IEM accuracy analyses. Smaller is more accurate. These are historical averages, not a forecast. As of 28 June 2026.
In 1907 the statistician Francis Galton described a weight judging competition at a fair in Plymouth, where visitors paid to guess the dressed weight of an ox. He gathered the entries, expecting the crowd to be wildly off. Instead, when he combined the 787 usable guesses, the average came to 1,197 pounds, within a single pound of the animal's true weight of 1,198 pounds, and closer than the typical expert in the room. He published the result in Nature under the title Vox Populi, the voice of the people.
The mechanism is plain once you see it. Each guess contains a mixture of signal and error. Some people guess too high, some too low, and the size of the mistakes varies. When the guesses are diverse and independent, those errors scatter in different directions and tend to cancel when you average them, leaving the shared signal to dominate. The crowd is not magic and it is not telepathy. It is statistics, the same reason a larger sample gives a steadier estimate than a single reading.
The story carries a warning as well as a promise. The averaging only works because the guesses were independent. Each visitor wrote a number without seeing the others. The moment people start copying a confident neighbour, the errors stop scattering and begin to line up, and the cancellation that made the crowd accurate quietly stops working. Hold on to that condition, because it is the one that fails most often in real markets.
A prediction market takes the averaging idea and adds two things a guessing jar lacks. First, it aggregates continuously through price, so the number updates the moment new information arrives rather than once at the end of a fair. Second, it attaches money to opinions. Because being right pays and being wrong costs, participants have a reason to do the work, to seek out information others have missed, and to trade against prices they believe are mistaken. That trading is the force that pushes a mispriced contract back toward a fair value.
The result is a price between one and ninety nine cents that reads directly as an implied probability. It is a weighted average in a real sense. Participants willing to stake more move it more, and those who are consistently right tend to accumulate the means to keep influencing it, while those who are consistently wrong run out of stake. This selection is why market prices are often a strong, hard to beat estimate of an uncertain event. It is also why beating them consistently is genuinely difficult, and why a feeling of easy edge is usually a warning sign rather than an opportunity.
It helps to be precise about what the price is and is not. It is the crowd's current opinion, expressed as a probability, weighted by conviction and money. It is not a forecast of the result, not a promise, and not a hidden message about what will happen. A contract at eighty cents is the market saying it would pay eighty cents for a dollar that arrives only if the event happens. It is a considered estimate that the event is more likely than not, and nothing stronger.
Collective judgement is accurate only under conditions, and each one can fail. Diversity fails when everyone shares the same blind spot, so the errors all lean the same way and there is nothing for the averaging to cancel. Independence fails when a single loud narrative takes hold and people stop reasoning for themselves. Aggregation fails when a market is thin, so a single large trader can shove the price around and the number stops reflecting a broad crowd at all. The incentive layer weakens when a venue has little volume or attention, because there is no one to trade against an obvious mistake.
Independence is the most fragile of the four, and the most worth watching. The very feature that makes a market powerful, a single visible price that everyone reacts to, is also what can undermine independence. When the price itself becomes the main piece of information people use, rather than their own research, the crowd starts to chase the number instead of informing it. That is the moment a wise crowd quietly turns into a herd, and the price can drift a long way from anything a careful analyst would call fair.
History is full of crowds that moved together off a cliff. Financial bubbles are the clearest case, where rising prices persuade more people to buy precisely because prices are rising, which is the exact opposite of independent judgement. The same dynamic can appear in any market where attention concentrates and reflection thins out. Prediction markets are not immune. A thinly traded contract, a viral but mistaken story, or a question where good information simply does not exist can all produce a price that looks confident and turns out to be wrong.
A confident price near one cent or near ninety nine cents is still a probability, not a certainty. Improbable things happen, and when they do they happen most painfully to people who treated a high price as a sure thing and staked accordingly. The honest reading is the one this whole page keeps returning to: the market price is a serious benchmark you should respect, and an estimate you should not worship. The wisdom of the crowd is a tendency, not a law.
This is the question people most want answered, and the honest reply is that the evidence is encouraging but contested. Published analyses of the Iowa Electronic Markets, a small academic exchange that has run election markets since 1988, report an average absolute error of about 1.34 percentage points for United States presidential vote share markets on election eve, and find the market closer to the final result than the polls roughly seventy four percent of the time, with the gap widest when forecasting many months ahead. Those are striking numbers and they are real.
They are also not the last word. Other researchers have looked again at the same comparison and found the edge smaller, or sensitive to which elections and which periods you include, and polling methods have changed over the decades the markets were tested against. The fair summary, marked here as contested rather than settled, is that markets and polls are both useful and imperfect, that markets often aggregate information well and update quickly, and that no source has earned the right to be treated as an oracle. A market price deserves a seat at the table, not the last word on the outcome.
Use the wisdom of crowds as a humbling default. Before you decide a price is wrong, ask why the aggregated judgement of many people with money on the line would have missed what you think you see. Sometimes the answer is genuine: you hold information the crowd lacks, or you are looking at a thin and inattentive market where the conditions for wisdom are not met. Often the honest answer is that the price is probably closer to right than your gut, and the feeling of edge is overconfidence wearing the costume of insight.
The practical habit is to check the conditions before you trust the price. Is the market liquid and widely watched, or thin and easy to push? Are participants reasoning independently, or all reacting to one story? Is there real information to aggregate, or is the question a coin flip dressed up as analysis? When the conditions hold, respect the price. When they do not, treat the number with the suspicion it deserves. And understand throughout that none of this is financial advice, that a market price can be wrong, and that the money you put at stake is genuinely at risk regardless of how wise the crowd appears.
One last discipline ties it together. The crowd is a benchmark to measure your own view against, not a verdict to copy. If your estimate and the price agree, you have learned that your thinking is in line with a large, motivated group, which is reassuring but not a green light. If they disagree, you have found a question worth investigating, not a guaranteed opportunity. Either way the useful move is to write down why you differ, in plain words, before any money is involved. A reason you can state and defend is worth more than a feeling of edge, and it is the only honest basis for treating a market price as wrong.
It helps to picture the same hundred people in two different rooms. In the first room, each person studies the question alone, writes down a probability, and hands it in without seeing anyone else's answer. Their mistakes are personal and uncorrelated, so when you average the hundred numbers the high guesses and the low guesses largely cancel, and the average is steadier and closer to the truth than almost any single entry. This is the wise crowd, and a deep, well traded market is the closest real thing to it.
In the second room, one confident person announces a number first, and the other ninety nine anchor on it before forming a view of their own. Now the answers are no longer independent. They cluster around the first voice, the errors line up instead of cancelling, and the average inherits whatever bias that first person carried. The room looks like a crowd and votes like one person. A thin market with a single dominant trader, or a market swept up in one viral story, behaves like the second room, and its confident looking price deserves far less trust than its precision suggests.
The difference between the two rooms is not the number of people, which is identical, but whether they reasoned independently. That is why headcount alone never tells you whether a price is wise. A small, diverse, independent set of informed traders can produce a better estimate than a huge crowd all reading the same feed. When you size up a market, you are really asking which room you are standing in.
A wise crowd can still be wrong, and a confident price is not a certainty. Prediction markets can lose you money. Stake only what you can afford to lose, never to chase a loss, and never on borrowed money. If it stops feeling like a free choice, step back. In the United States you can call or text the helpline on 1-800-GAMBLER or visit ncpgambling.org.
The wisdom of a crowd depends on how many people are in it and how freely they trade, which is to say on liquidity and fees. If you want to see how deep and how costly the regulated venues are before you read a price as a crowd's verdict, our cross platform data is the place to start. Availability and eligibility vary by region, so confirm your own access before opening any account.
It is the finding that the combined judgement of many independent people can be more accurate than most individuals within the group, and sometimes more accurate than experts. The classic illustration is Francis Galton's 1907 account of a fair where the average of 787 guesses at an ox's weight, 1,197 pounds, came within one pound of the true 1,198 pounds.
A market price aggregates the views of many participants who are each putting money behind their opinion. Trading rewards being right and penalises being wrong, which gives people a reason to research and to correct mispricing. The price that results is a weighted summary of the crowd's current view, expressed as an implied probability.
James Surowiecki's account points to four conditions: diversity of opinion, independence so people are not copying each other, decentralization that lets local knowledge in, and a mechanism to aggregate the views. A market supplies aggregation through price and adds honest financial incentives. When these hold, errors tend to cancel; when they fail, the crowd can be biased or herd.
Sometimes, and the question is debated. Published analyses of the Iowa Electronic Markets report an average absolute election eve error of about 1.34 percentage points for US presidential vote share markets, and that the market beat the polls roughly 74 percent of the time. Other research finds the edge smaller or sensitive to the period studied, so treat the market as a strong benchmark rather than a settled winner.
Yes, often and sometimes badly. When independence breaks down and people copy one another, when information is poor, or when a market is thin and easily pushed, the crowd can be confidently wrong. A market price is current opinion, not a forecast of the result, and it can move sharply.
Treat it as a serious benchmark, not as truth. The price is a useful aggregate that is hard to beat, but it is still an estimate that can be wrong. We never name a predicted winner, and a price near certainty is not a guarantee.
The rules change fast. Get the changes that affect you, plain and current, not tips.
Independent. Every claim dated and sourced. No platform pays for its place.