Does 65¢ really mean a 65% chance?

The useful shortcut

For a contract paying $1 on success and $0 on failure, a price of 65¢ is commonly described as a 65% implied probability. Divide the price by the winning payout to get that number. This convention describes a market price; it does not establish the true probability of the event.

The arithmetic behind it

Imagine a hypothetical outcome with a 65% chance of paying $1. Its expected payout is 0.65 × $1 + 0.35 × $0 = $0.65. A risk-neutral buyer paying 65¢ would break even in expectation before fees and the cost of tying up money.

Expected value is an average across hypothetical repetitions. A single contract still resolves to its specified payoff. A positive expected value does not prevent a loss.

A price is not an average vote

Participants differ in their beliefs, resources, and willingness to take risk. The people who trade are not a representative survey of everyone who follows an event. A price alone cannot tell you how widely traders disagree.

Manski shows that, under his model of risk-neutral traders with differing beliefs, a price does not reveal the dispersion of those beliefs and only partially identifies their central tendency. Wolfers and Zitzewitz derive conditions under which prices equal mean beliefs and find that prices often approximate average beliefs in a broader class of models. Neither result makes a market price a verified real-world probability.

Check which price you are reading

A chart may show the last trade, a midpoint, or another reference price. These can differ from the price available to a new buyer. Polymarket’s international documentation, for example, describes displaying the midpoint, with the last traded price shown when the bid–ask spread exceeds 10¢.

If the best bid is 62¢ and the best ask is 68¢, the midpoint is 65¢. That does not mean you can buy at 65¢. A useful quotation states the venue, contract, timestamp, and whether it uses a bid, ask, midpoint, or last trade.

Unlikely outcomes still happen

An event assigned a 10% chance can occur without proving the forecast was unreasonable. Calibration is evaluated over many forecasts: events forecast near 10% should happen about one time in ten in a well-calibrated set. One dramatic result cannot establish a platform’s overall accuracy.