What Is a Prediction Market?

A prediction market is a market in contracts whose payoff depends on whether a defined event happens. Buy YES at 37 cents and you are paying 37 cents for a claim that pays a fixed amount — usually one dollar — if the event described in the contract rules occurs, and nothing if it does not. The price you see is the number at which buyers and sellers are currently willing to trade that payoff. It is a price, and only a price. Everything useful about prediction markets, and every way of misreading them, follows from that one sentence.

The contract is the thing, not the headline. A market title is a summary. The contract rules are the product. They name the condition that counts as YES, the deadline, the time zone, the source the outcome will be read from, and what happens in the messy cases nobody thinks about until they arrive. Two markets can carry near-identical titles and near-identical prices and still be different contracts, because one of those rules differs. Price proximity is not equivalence. If you are comparing two venues on what looks like the same question, read the rules on both before you compare the numbers at all.

What the price is, and what it is not. Because a YES contract pays one dollar, a price of 37 cents maps neatly onto “37 percent”. Venues make that mapping explicitly, describing the price as the market-implied probability of the outcome. The mapping is genuinely useful. It is also an interpretation that holds only under assumptions: that fees and spreads are small relative to the number you care about, that money tied up until settlement costs nothing, and that enough participants are trading for the price to mean anything at all. When those assumptions fail, the number is still a price. It just stops being a good probability estimate.

There is a second gap: the price on the screen is usually not the price you can trade. A displayed quote is typically derived — a midpoint between the best bid and the best ask, or a last trade when the spread gets wide. Buying, you pay the ask. Selling, you receive the bid. The difference is small in a busy market and embarrassing in a quiet one.

Someone has to decide the outcome. A contract only pays once the outcome is settled, and settlement is a process with an owner. On a regulated exchange that owner is the exchange, operating under a published rulebook. Elsewhere it can be an oracle process in which anyone may propose an outcome, anyone may dispute it, and a disputed case escalates to a vote on a timetable measured in days. Neither arrangement is automatically better, but they fail differently.

How this differs from the nearby things. A poll samples stated intentions and reports them with a margin of error. A model forecast applies stated assumptions to data and reports an output. A prediction market reports what people will pay, right now, for a defined payoff. Whether that makes it more accurate than a poll is a question this guide does not answer, because we have not sourced an answer. The CFTC tells customers to check costs, and warns that trading with unregistered entities may leave them with little or no protection.

The three misreadings worth naming. First: treating a price as a verdict. A market at 80 cents has not decided anything; it has priced something. Second: treating one resolved market as a scorecard. A well-calibrated 70 percent should be wrong three times in ten. Third: treating a similar price on two venues as the same bet. It is the same bet only if the contract terms and payoff match.

A reading routine. Read the resolution rules before the price. Check which price you are looking at, and what you would actually pay to trade it. Check whether anyone is trading at all. Note who resolves the contract and from what source. Then treat the price as one estimate among several, and say out loud what would have to be true for it to be wrong.