Behavioral biases can make unlikely outcomes too expensive, slow a market's response to unwelcome news, or amplify buying after a widely shared headline. They affect prices when traders act on those judgments and other participants do not fully offset their orders.
For someone reading market odds, the consequence is straightforward: a displayed probability can reflect both useful information and systematic errors in how people interpret it. The size and direction of those errors vary by market. Research has found longshot overpricing in some settings, no such pattern in others, and even cases where changing the way outcomes are grouped changes their prices.
How a trader's bias reaches the market price
A Yes contract that pays $1 if an event happens and $0 otherwise is commonly interpreted as a probability: a 60¢ price implies roughly 60%. That price-to-probability conversion makes prediction markets useful as forecasts.
But prices come from orders. If optimistic buyers keep accepting higher offers, the quote can rise even without new evidence. How far it moves depends on the shares available at each price and the traders willing to sell.
A biased belief matters less when its holder never trades. A smaller group with substantial buying power can matter much more. Wolfers and Zitzewitz's analysis of market prices explains why prices can approximate average beliefs while still depending on risk preferences and the distribution of those beliefs.
Favorite–longshot bias: when low odds cost too much
A 5¢ share looks cheap. If it wins, it pays $1. The question is whether its chance of winning justifies the price.
Suppose your probability estimate is 2%. For one share held to settlement, before fees:
Expected profit = (0.02 × $1) − $0.05 = −$0.03
Your estimate may be wrong, but a large potential payout does not change that calculation. Overweighting a small chance of winning can make a longshot attractive at a price its probability does not support.
The favorite–longshot bias describes a recurring pattern of relatively poor returns on unlikely outcomes compared with favorites. A September 2026 Polymarket preprint by Marcos Cardozo and José Ignacio Rivero-Wildemauwe used a dataset covering 588 million trades. Comparing purchase prices with settlement payouts, the authors found that purchases below 10¢ lost 6.3% on average when each individual market received equal weight. When related markets were grouped under their parent event, such as candidates in one election, and each event received equal weight, the result became a 4.1% gain.
Those calculations exclude fees and assume holding to settlement; traders who sold earlier could have earned different returns. The working paper documents a pricing pattern without establishing one psychological cause. Even the sign of the result depends on what receives equal weight in the sample.
Nor is longshot overpricing universal. Berg and Rietz's study of the Iowa Electronic Market found no longshot bias in the markets they examined. A low price warrants a probability calculation, not an automatic bet against the outcome.
Confirmation bias and overconfidence
Imagine holding Yes on a candidate you support. You accept a favorable poll immediately but spend an hour searching for flaws in an unfavorable one. The issue is the different standard of evidence. This is confirmation bias, and it can make traders slow to revise a forecast when the facts turn against their position.
Overconfidence can make the problem more expensive. A trader who treats a fragile estimate as nearly certain may place a larger order or accept a worse price. If other participants share the same conviction, demand can remain strong despite contradictory news.
There is documented evidence that optimism can reach prediction market prices. In a 2008 study of Google's internal markets, researchers reported that contracts tied to favorable company outcomes were overpriced by about 10 percentage points during their study period. Longer-tenured employees and more experienced traders were better calibrated. These markets used virtual currency, so the size of the effect is specific to that setting.
For your own forecast, decide in advance which evidence would make you change it. If several unfavorable polls would count as noise but one favorable poll would justify buying more, the problem is already visible in your decision rule.
Herding: when the price becomes its own evidence
Consider a hypothetical contract on whether a company will launch a product by December. A report says testing is ahead of schedule. Social accounts repeat it, the Yes price rises, and screenshots of the rising odds attract more buyers.
Someone arriving late sees a report, apparent agreement, and a price move. All three may come from one piece of information. Counting them separately exaggerates the case for Yes.
Herding can amplify that feedback: traders follow the buying because other buyers appear informed. Repeated coverage can also make the story easier to recall and give it disproportionate weight, a mechanism associated with the availability heuristic.
Following other traders can be sensible when they know something you do not. To assess this particular move, trace the report to its original source and check what changed. Testing ahead of schedule may matter, but it does not establish whether the contract's definition of a launch will be met. A limited beta and general availability could produce different settlements.
Rising price and volume alone cannot tell you whether the market is learning or copying.
Framing: how grouping outcomes can change their prices
Prediction markets can present the same uncertainty in different ways. A market on total goals might group outcomes into three ranges: 0–1, 2–3, and 4 or more. Another could offer five: 0, 1, 2, 3, and 4 or more.
The chance of four or more goals has not changed. Yet people may start by giving each displayed category similar weight, then adjust too little for what they know. The grouping becomes part of the forecast.
This effect, called partition dependence, has been tested in markets. Sonnemann and colleagues found evidence in controlled laboratory trading and a field experiment involving NBA and World Cup events. Their work shows that a bias in individual probability judgment can survive trading and affect market prices.
The goals example above illustrates the mechanism. When comparing markets with different outcome lists, check that you are comparing equivalent events and consider how the grouping directs attention.
Anchoring and loss aversion after a price falls
Suppose you buy Yes at 65¢. Bad news arrives, the price drops to 35¢, and you lower your forecast. You still decide to wait for 65¢ so you can break even.
That entry price has become an anchor. Loss aversion can make selling below it feel particularly uncomfortable. The related disposition effect describes the tendency to hold losing investments too long and sell winners too readily.
These habits can make traders reluctant to revise their orders after news. Their effect on the overall price depends on the other buyers and sellers; one stubborn holder does not establish a market-wide error.
For an event contract, waiting ends at settlement. Reassess the position using your current probability estimate, the price you could actually sell at, and the costs of doing so. Your entry price records what you paid. It does not tell you what the contract is worth today.
Why competing traders do not remove every bias
If a contract looks underpriced, other traders have a reason to buy it. That can correct an error, but the trade requires money, time, and exposure to an uncertain outcome.
A 98¢ contract paying $1 offers only 2¢ of profit if it wins and a 98¢ loss if it loses, before costs. Even a trader who believes it is underpriced may be unwilling to commit much capital for several months.
Time to settlement also affects how prices should be interpreted. Page and Clemen found stronger favorite–longshot miscalibration farther from expiration in their dataset. Their model links that pattern to the cost of tying up money. An apparent probability error can therefore have an economic explanation as well as a psychological one.
How much should you trust the odds?
Use the price as a forecast you can investigate. Read the settlement terms, check the available bid and ask, and identify the evidence behind a move. A stale last trade, different deadlines, or a wide spread can explain a suspicious-looking quote without a behavioral story.
An unexpected winner is not proof of bias. In a well-calibrated market, outcomes priced around 20% should happen roughly one time in five across many comparable forecasts. Assessing accuracy requires a series of observations made at comparable points before settlement; a Brier score can help compare those forecasts.
If you disagree with the market, record your estimate alongside the available quote and the evidence for the difference. After a series of markets settles, compare your forecasts with those prices at the same timestamps. That gives you a way to check whether your adjustments improved accuracy.
Artem Goryushin
Fintech expert, business analyst
Artem is a fintech expert and business analyst with experience in prediction markets and financial analytics.