Back to blog

UNDER THE HOOD

Who Decides That YES Won

The market closes — but who says how it ended? The answer doesn't come from Polymarket. It comes from an oracle, a posted bond, and people who lose money when they lie. Let's look at it up close.

The market runs, the price moves, the deadline passes. And then comes the moment almost nobody thinks through in advance: somebody has to say how it turned out. Not guess, not comment — decide, bindingly, who gets a dollar a share and who ends up with nothing.

It won't be a Polymarket employee glancing at the news and clicking a button. If it were, the whole market would stand or fall on trusting one company — which is precisely what this world is trying to avoid. What decides is a mechanism called an oracle, and it rests on one idea: lying has to cost more than staying quiet.

The oracle: a door between the world and the chain

A blockchain is a closed system. It can remember who sent what to whom, but it has no way of finding out whether it rained outside or who won an election. An oracleThe mechanism that gets a piece of the real world into a smart contract. The contract can't verify it, so the credibility of the whole outcome stands or falls with how the oracle is built. is the service that carries outside information in — and it's the most fragile part of the whole structure.

Polymarket uses what's called an optimistic oracle. The name fits exactly: the system assumes people are telling the truth and only verifies when somebody raises a hand. The sequence is simple:

  1. After the deadline, anyone can propose an outcome and post a bond alongside it — hundreds of dollars, in that order of magnitude.
  2. A challenge window runs, a matter of hours. If nobody objects, the proposal becomes the outcome and the proposer gets the bond back plus a reward.
  3. If somebody disputes it, they have to post a bond of their own, the same size. That sends the dispute to a vote — and whoever turns out to be wrong loses their bond to the other side.

It sounds almost banal until you try out what happens in each of the combinations. Play both roles:

Resolution simulator three steps, four endings

First, set reality: how did it actually turn out?

The numbers are indicative — the exact bond size and window length change over time. The principle stays: both sides of a dispute have their own money on the line.

Why lying usually doesn't pay

Notice what the simulation shows. Proposing the truth is almost risk-free. Proposing a lie means betting that nobody looks for two hours — and if somebody does, you lose your bond. Disputing blindly is the same trap in reverse. So the system doesn't pay for truth; it just makes lying unpleasantly expensive and relies on somebody watching the profitable markets. Usually somebody is — and it's people with their own money in that market.

Who watches the watchmen When it comes to a vote, the holders of that oracle's token decide. They have an incentive to vote correctly, because their token's value rests on the system's reputation. It's the best available solution — not a perfect one. It's a vote, not a court.

And that's also the main vulnerability. Once there's more money in a market than there is in the bonds and voting tokens combined, attacking starts to pay: you can lose on bonds and win on the position. Which is why on huge, contentious markets it's worth looking not only at the price, but at how clearly the question is written.

The wording decides, not the truth

Nine disputes out of ten don't happen because somebody lied. They happen because the question allowed two readings. A market isn't resolved by what happened, but by what the resolution rules say — the text tucked behind a link on the market page that almost nobody reads.

See whether you can spot a question that won't survive a dispute:

Will this question hold up? 4 wordings
  • Will the S&P 500 close above 6,000 points on 30 September 2026, per the exchange’s official closing value?

  • Will there be peace in the Middle East by the end of the year?

  • Will the manufacturer announce a new model by 30 June?

  • Will party X get the most votes nationwide according to the official results of the electoral commission?

Nothing yet — take the first one.

A good question always has three things: a deadline including the time zone, a threshold or exact condition, and a source the answer is read from. If one of them is missing, it isn't a market about the future — it's a market about how it'll be interpreted.

The end of the market isn't the end of the bet Between the deadline expiring and the payout you'll wait anything from hours to days — and in a dispute, longer. During that time the money isn't yours and isn't anybody else's; it just waits. Bear that in mind when you're betting on something that ends right before you need the cash.

What to do with this when you bet

Read the resolution rules before the price. The cheapest way to lose a sure bet is to be right about the world and wrong about how the question was written.

Ask who the source is. When the source is an official body, an exchange or a statistical office, a dispute is unlikely. When the source is "widely available reporting", expect a vote to settle it in the end.

A price near the edge isn't free money. A share at 97 cents the day after a market closes looks like a three-percent certainty. What you're actually paying for is the risk that the resolution lands differently than it seems — and that's the one risk the market hasn't closed yet.

What to take away

  • The operator doesn't decide the outcome, an oracle does — a mechanism where both sides of a dispute post their own money.
  • The optimistic model approves a proposal by silence. Only what somebody disputes ever gets verified.
  • Lying doesn't pay as long as the bond costs more than the position gains. On huge markets, that equation stops holding.
  • Most disputes are about wording, not facts. Deadline, threshold, source — without those three it isn't a question.
  • Read the resolution rules before you look at the price.

You can read a market, you know the route the money takes, and you know who decides. That leaves the least romantic and by far the most important part: how much to actually put into a single bet.