Stellaris

Learn

How to read prediction-market odds

A prediction-market price reads like a probability, and it mostly is one. What it does not tell you is how much money stands behind it. Four quick checks, and a worked example.

Last reviewed 17 November 2026 · Method v1

The short version

A prediction-market price is a number between 0 and 100 cents, and you can read it as a probability: 70 cents means the market puts about a 70% chance on yes. That is true as far as it goes.

What the number does not show is how much money stands behind it, how easily one trader could move it, or whether the other venue agrees. Two prices of 70% can mean very different things. This guide covers what the number is, then four checks that tell you how far to lean on it.

What the number actually is

On Polymarket, each share pays $1 if the outcome happens and $0 if it does not, so the price is a probability in dollars. The price shown on the page is not a trade. It is the midpoint between the best buy offer (the bid) and the best sell offer (the ask). If the gap between them is wider than 10 cents, Polymarket shows the last traded price instead.

That has two consequences. You will not trade at the shown price: you pay the ask when you buy and receive the bid when you sell. And on a wide-spread market the number on the page can be stale, because it is the last trade, not a live quote.

What 70% means, and does not mean

A price of 70% is not a forecast that something will happen. It says the event should happen about seven times in ten. In the week Nvidia's chief executive stepped down, the market on his departure sat at 75 to 80% for fourteen hours before it happened. The market was saying "likely", not "done", and that was the right thing to say.

Whether prices like that are calibrated depends heavily on the money behind them. When we scored 316 midterm markets, favourites won 91% of the time when more than 100,000 contracts rested near the price, and about 73% when fewer than 10,000 did.

Four checks before you lean on a number

  • Spread. The gap between the best buy and sell price. One cent or less is a tight market. Ten cents or more means the price is closer to a guess than a quote.
  • Depth. The number of contracts resting within five cents of the price. In our data, markets under 1,000 contracts are thin, and 10,000 or more are deep. Depth is what stops one trader moving the price.
  • Agreement. Does the same question on the other venue say roughly the same thing? If Polymarket and Kalshi are 20 points apart on one question, check the rules before you read either number.
  • Does it hold? Look at the price an hour later. In a week of 576 large moves, moves on thin books gave back a median 30% of the move within an hour. Moves on deep books gave back 9%.

A worked example: the December Fed meeting

On Sunday 15 November, Polymarket priced a 25 basis point rate hike at 86% for the December meeting, and Kalshi at 87%. Run the checks. The spread on both was a cent or two. Polymarket's market had about $34 million traded, and Kalshi had more than 1.9 million contracts open on the 25 basis point hike. The two venues were within a point of each other, and the price had barely moved for days. That is a number you can lean on.

Compare a different market that week: the race to be Nvidia's next chief executive. Soon after it opened, one candidate traded at 48%, in a market with about $14,000 traded on him in total. Six hours later he was back near 8%. The number looked just as precise. It was a handful of orders.

Two traps worth knowing

Deadline questions. A question that pays yes if something happens any time before a date, such as "Will oil hit $100 by December?", behaves differently from a question settled on one day. In our earlier study, middle prices on "hit X by a date" markets overstated the chance of yes, most of all close to the deadline.

Outcomes that should add to 100%. When a question has several outcomes and only one can win, their prices should add to about 100%. If they add to much more, someone's quote is stale or one outcome is thinly traded. Add them up before you trust any one of them.

The checklist

  • Is the spread a cent or two?
  • Are there thousands of contracts resting near the price, not hundreds?
  • Does the other venue agree, on a question with the same rules?
  • Is the price still there an hour later?
  • If it is a deadline question, have you allowed for the overstatement?

A price that passes all five is a reasonable read of what informed traders think. One that fails two or three is colour, not a probability. Neither is advice, and neither tells you what will happen.

Keep reading

Sources

This is analysis of public market data, not investment advice or a recommendation to trade. Stellaris is not a trading venue, broker or adviser and does not trade. Prices are those shown by the venues at the times stated and can change. Trust levels come from our published method and have known limits.

See the latest briefRequest access