Updated July 2026

How to read prediction market prices

A prediction market price is a probability wearing a dollar sign. Once you can translate one into the other without thinking, every market page stops being a betting slip and becomes a statement you can agree or disagree with. This guide covers the full reading craft: cents to implied probability and back, the overround, how these prices differ from bookmaker odds, what actually moves them, and the mistakes that trip up almost every new reader.

A price in cents is a probability in disguise

Every binary market quotes two instruments: a YES share and a NO share. Each pays exactly one dollar if its side turns out correct and nothing if it does not. That fixed payout is what makes the price readable. If a YES share trades at 58 cents, buyers as a group are paying 58 cents for a claim on one dollar, and that only makes sense if the crowd believes the event happens about 58 percent of the time. This is the prediction market odds meaning in one line: the price in cents is the market's implied probability in percent.

The two sides are mirror images. If YES trades at 58 cents, NO should trade near 42 cents, because holding one share of each always pays exactly one dollar at resolution no matter which way the event goes. Whenever YES plus NO drifts meaningfully away from a dollar, arbitrage traders buy the cheap pair or sell the expensive one and push it back. That is why on liquid markets you will almost always see the pair sum to roughly one dollar, give or take the spread.

Converting prices to probabilities and back

The conversion is deliberately trivial, and it runs both directions:

• Price to probability. Divide the price in cents by 100. A YES at 27 cents implies a 27 percent chance. A YES at 91 cents implies 91 percent.
• Probability to price. Multiply your own forecast by 100. If your research says an event is 70 percent likely, your fair value for YES is 70 cents and your fair value for NO is 30 cents.

The payoff math follows directly. Buy YES at 58 cents and the event happens: the share redeems for one dollar, a gain of 42 cents on 58 risked, roughly a 72 percent return. The event does not happen: you lose the full 58 cents. Run those two branches at any assumed true probability and you get the single most useful sentence in this whole subject: buying at 58 cents makes sense in expectation only if you believe the true probability is above 58 percent. Everything else on this site is an elaboration of that sentence.

One more worked example from the NO side, because beginners often forget it exists. Suppose YES on some longshot trades at 12 cents but your base rate work says the real chance is closer to 4 percent. You do not need to touch YES at all. NO is priced at about 88 cents while your model says it is worth 96. Buying NO at 88 risks 88 cents to win 12, and if your 96 percent figure is right, that trade wins roughly 24 times out of 25.

Who decided the price is 58?

Nobody set it. A prediction market runs an order book, the same machinery as a stock exchange. The last trade at 58 cents happened because someone who thought the probability was higher than 58 met someone who thought it was lower, and they transacted. The displayed price is simply the most recent point where an optimist and a pessimist agreed to disagree with money attached.

That makes the price a money-weighted opinion poll. People who feel strongly and are willing to risk more move it further than people posting takes for free. It does not make the price an oracle. It means the price summarizes what informed and semi-informed participants collectively believe right now, which is a useful baseline and nothing more.

The overround: why the numbers can sum past 100

Look closely at a live book and you will notice the tradeable prices do not sum to exactly one dollar. You might see YES offered at 59 cents and NO offered at 43 cents. Buy both and you pay 102 cents for a package worth exactly 100 at resolution. That extra 2 cents is the overround, and in an order book market it is just the bid-ask spread wearing another name. It is the cost of demanding immediate execution rather than posting your own order and waiting.

The practical fix: read the midpoint, not the ask. If YES is bid 57 and offered 59, the market's real estimate is about 58 percent, and the extra cent on either side is a transaction cost, not information.

Multi-outcome markets make the overround more visible. Take a three-candidate race quoted at 52, 33, and 19 cents. Those sum to 104, and probabilities cannot sum to 104 percent. To recover honest implied probabilities, normalize: divide each price by the total. That gives 50 percent, 31.7 percent, and 18.3 percent. Skipping this step quietly inflates every candidate's chance, and the inflation is worst exactly where beginners look for longshots.

How this differs from bookmaker odds

Bookmaker odds encode the same idea with the house's margin baked directly into the number. Decimal odds convert as one divided by the odds: a line of 1.72 implies 58.1 percent. American odds of -140 imply 140 divided by 240, about 58.3 percent. So far, so similar.

The difference appears when you price both sides. A bookmaker might offer 1.72 on one team and 2.10 on the other. Convert both: 58.1 percent plus 47.6 percent is 105.7 percent. That 5.7 points is vig, the bookmaker's fee, spread invisibly across both prices, and it does not move to zero when you trade patiently. On a prediction market the equivalent cost lives in the spread instead, and it shrinks toward nothing on liquid markets or when you post limit orders. That is why the midpoint of a deep prediction market is generally a cleaner probability estimate than a bookmaker line: there is no vig baked the same way into the quote itself. The market can still be wrong, but it is wrong honestly.

What moves a price

A price only changes when someone trades, so every move is one of three things:

• News. A poll drops, a candidate withdraws, a court rules, an earnings number prints. Fast repricing on real information is the market working as designed, and liquid markets often move within seconds of a headline.
• Flows. A large order walks through the book and moves the price several cents without any new fact in the world. On thin markets a few hundred dollars can shift the displayed price 10 points. Flow moves tend to partially revert as other traders fade them.
• Whales and smart money. Some wallets have long public track records, and other traders follow their entries, so a single large, identifiable position can move a price twice: once mechanically and once through imitation. Whether that move carries information depends entirely on whose money it is.

There is also a slow structural drift: as resolution approaches, prices get pushed toward 0 or 100 because uncertainty is running out of time to matter. A market sliding from 65 to 72 over three quiet weeks may reflect nothing but the calendar.

When the price disagrees with the evidence

This is the core PredictionSignal idea. A price is a claim, and claims can be checked. Suppose a market prices an incumbent's re-election at 58 cents, but incumbents in comparable races with this polling position have won about 72 percent of the time across several decades of data. That 14 point gap is the raw material of a signal. Before trusting it, run the checklist that separates analysis from wishful thinking:

• Read the resolution rules. Many "obvious" mispricings dissolve when you learn the market settles on a technicality you skimmed past.
• Assume the market knows something first. Search for the news you might have missed. Sometimes the price is early, not wrong.
• Check depth and volume. A stale price on a thin book is not a considered market opinion, it is an artifact of nobody trading.
• Then, and only then, treat the gap as a candidate edge. Our methodology page shows how we formalize exactly this comparison across a panel of AI analysts.

You can practice the reading loop on any live venue. Open a market on Polymarket, write down the implied probability from the midpoint before reading any comments, then go looking for the base rate. The habit of forming your own number first is worth more than any single trade.

Common price-reading mistakes

Most losses attributed to bad luck are actually bad reading. The recurring offenders:

• Treating 90 percent as certain. A 90 cent favorite fails one time in ten. If you buy ten such markets, expect one to blow up, and expect the loss on that one to erase most of the small wins. High prices are not safe, they are asymmetric.
• Ignoring time to resolution. A YES at 60 cents resolving next week and a YES at 60 cents resolving in fourteen months are completely different trades. The second locks your capital, exposes you to a year of unknown news, and pays the same 40 cents. Always read the date next to the price.
• Reading last-trade price on thin books. The big number on the page is often the last trade, which on a quiet market may be days old. Check the actual bid and ask. If they are 44 and 71, the market has no opinion worth quoting, whatever the headline number says.
• Skipping normalization in multi-outcome markets. As shown above, unnormalized longshot prices overstate every outcome's chance. Divide by the sum before comparing to your own estimate.
• Confusing movement with information. A price that jumped 8 points demands an explanation, not a chase. If you cannot find the news, you may be looking at flow, and flow reverts more often than it continues.

None of this makes prediction markets unreadable. It makes them readable the way any market is: carefully, with the mechanics in view, and with the humility to remember that a probability is a long-run statement, not a verdict on the next single outcome. This guide is education, not financial advice, and no method of reading prices removes the risk of being wrong.