How to read a prediction market price: 62¢ is a probability

You are looking at a market that says 62¢. Most people read that as "the market predicts yes." It does not. A price of 62¢ is the market's estimate that the event has about a 62% chance of happening, so it expects "no" 38% of the time.

This post covers how to read that number, what the studies say, and how much to stake.

What does a price of 62¢ actually mean?

A binary contract pays 100¢ if the event happens and 0¢ if it does not. If it trades at 62¢, buyers are collectively willing to pay 62¢ for a 100¢ payout, which implies a probability near 62%.

The "implies a probability" step needs care. Economists Justin Wolfers and Eric Zitzewitz set out the conditions under which prices correspond to the average belief of traders in a May 2006 NBER working paper, responding to an earlier objection that there was little theory behind reading binary prices as probabilities. Their conclusion: prices give useful, though sometimes biased, estimates of the average belief.

What does calibration mean?

A forecaster is calibrated if events given a 62% chance happen about 62% of the time across many cases. Take every market that traded near 60¢ and count the yes results. About 6 in 10 means calibrated, 7 in 10 means underconfident, 5 in 10 means overconfident.

One resolved market therefore tells you little about whether a price was good. A 30% event happening is not an error.

What does the published evidence say?

The oldest evidence comes from the Iowa Electronic Markets, small real-money election markets run by the University of Iowa. In a paper in the International Journal of Forecasting, Joyce Berg, Forrest Nelson and Thomas Rietz compared the markets to 964 polls across five US presidential elections from 1988 to 2004. The market was closer to the eventual two-party vote split 74% of the time, and when forecasting more than 100 days ahead it significantly outperformed the polls in every election. A 2008 chapter with Robert Forsythe reviewed 49 markets across 41 elections in 13 countries, and prediction markets vs polls covers the polling side.

More recent work looks at the large commercial platforms. A February 2026 preprint by Nam Anh Le, revised on 4 August 2026, analysed 353 million trades across 429,000 binary contracts on Kalshi and Polymarket. Its most consistent finding is underconfidence in political markets, where prices sit closer to 50% than outcomes justify. The author's summary: "calibration is therefore conditional", depending on domain, time to resolution and trade size.

Note the limits. Iowa's markets are small and cover mostly elections, and the newer studies are working papers or preprints, not settled results.

What is the favourite-longshot bias?

The best documented pricing quirk is the favourite-longshot bias. Longshots (low prices) tend to win less often than their price implies, and favourites (high prices) tend to win more often. It was first studied in horse racing, where betting odds overprice longshots and underprice favourites, as Erik Snowberg and Justin Wolfers described in an April 2010 NBER paper (later published in the Journal of Political Economy). They concluded that misperceived probabilities, as in prospect theory, explain it better than risk-loving gamblers.

The same shape has since been found in a prediction market. In a working paper titled "Makers and Takers: The Economics of the Kalshi Prediction Market", Constantin Bürgi, Wanying Deng and Karl Whelan used transaction data on more than 300,000 Kalshi contracts. They report that prices become more accurate as markets approach their close, that low-priced contracts win far less often than their prices require to break even, and that high-priced contracts win more often and earn small positive returns.

The practical reading is modest. A 4¢ contract is often priced a bit higher than its true chance, and a 94¢ contract is often priced a bit lower. Treat it as a tilt, not a rule to trade blindly.

Why can a market at 95¢ still lose?

Because 95¢ means 5 times in 100, not never. Across 100 perfectly calibrated 95¢ markets, about 5 resolve no, and any single one could be among them.

There is also a second way to lose that has nothing to do with probability: resolution. A market resolves according to its written rules and its named source, not according to what you meant by the question. If you have not read the criteria, you may be pricing a different event than the one that settles. See who decides what actually happened.

How do you compare a price to your own estimate?

The number that matters is the gap between the price and what you honestly think the probability is. If you think an event is 70% likely and the price is 62¢, you see a positive expected value. If you think it is 55% likely, the same price is a bad buy, and the better side of the trade is the other one.

The arithmetic is simple. Buying a yes share at price p, with your own probability estimate q, has an expected profit per share of q minus p (using a 0 to 1 scale). At 62¢ and 70%, that is 8¢ per share on average, and you still lose that trade 3 times in 10.

The weak point is q. It is your guess, and a gap can mean you have an edge or that you are missing something. Ask what would need to be true for the market to be right. Often you know nothing the crowd does not, and passing is a valid trade.

How much should you stake on one idea?

Position sizing is where an edge is either kept or lost. The standard framework is the Kelly criterion. Edward Thorp's paper, presented in 1997 and revised in 1998, describes it as maximising the expected value of the logarithm of wealth. In plain terms, it sizes each bet to grow a bankroll fastest over many bets without wiping it out.

For a binary contract bought at price p with your estimated probability q, the Kelly fraction of your bankroll is (q minus p) divided by (1 minus p). At 62¢ with a 70% estimate, that is 0.08 divided by 0.38, or about 21% of your bankroll. It assumes your 70% is exact.

Because your estimate is noisy, a stake that large is rarely wise. Treat the formula as a ceiling, not a target, and spread your Sense across many small positions so no single resolution decides your balance. New accounts start with 1,000 Sense, enough for dozens of small trades (first prediction guide).

Ravioli is a free-to-play prediction market. You trade with Sense, which is free, and the prizes are real.

Create a free account and browse the board.