How Do Prediction Markets Work?

A prediction market lets you buy and sell contracts that pay out based on whether a real-world event happens. Will the Fed cut rates in September? Will a given team win the title? Will a candidate take a state? Each question becomes a tradable contract whose price moves as new money and new information arrive. The two venues most US traders use are Kalshi, a CFTC-regulated exchange, and Polymarket, which runs on-chain. The mechanics below apply to both.

YES and NO contracts priced 0 to 1

Every market is built from binary contracts. A YES contract pays $1 if the event resolves true and $0 if it does not. A NO contract is the mirror: $1 if the event is false, $0 if it is true. Because exactly one side wins, the two prices are tied together by a hard constraint.

YES price + NO price = $1.00 (roughly, before fees and spread).

So if YES trades at 65 cents, NO trades near 35 cents. Buying NO at 9.5 cents is mathematically the same trade as selling YES at 90.5 cents. This identity is why arbitrageurs watch both legs and both venues. If the two sides drift apart enough to cover fees, that gap is free money until someone closes it.

Price is the market's probability estimate

Here is the part that makes prediction markets useful: the price of a YES contract is the market's consensus probability of the event, expressed as a number between 0 and 1. A contract trading at 65 cents means the market collectively prices the outcome at about 65 percent.

The logic is simple expected value. If you think the true probability is 75 percent and YES is trading at 65 cents, you are getting a $1 payoff for 65 cents on something you believe hits three times out of four. Positive edge. You buy. Other traders who disagree sell to you. The clearing price is where the marginal buyer and marginal seller agree, which is the crowd's best estimate at that moment. When fresh information lands, traders reprice and the implied probability moves with it.

The order book: bids, asks, and the spread

Both Kalshi and Polymarket run a central limit order book, a continuous double auction. Two stacks of resting orders define the market:

• Bids are what buyers will pay. If the best bid on YES is 62 cents, someone has a standing order to buy there.
• Asks are what sellers will accept. If the best ask is 65 cents, that is the cheapest price you can buy right now.
• The spread is the gap between them, 3 cents in this example. It is the cost of immediate execution.
• Depth is how many contracts sit at each price level. It decides what large orders actually fill at.

Tight spreads of 1 to 2 cents mean a liquid, well-traded market. Spreads of 5 cents or more signal thin liquidity, genuine uncertainty, or both. A market order fills against the resting orders by price-time priority: best price first, earliest order breaks ties. Want 500 contracts at 62 cents but only 200 are resting there? You fill 200 at 62, then either climb the book to worse prices or, with a limit order, leave the remaining 300 waiting.

This is the single most common beginner mistake. The price on a market card is the last trade, not your trade. For anything beyond a small order, read the book depth, not the headline number.

A concrete example

Suppose a market asks: will a specific company IPO before year-end? The book shows YES bid 90.0 / ask 90.9, NO bid 9.1 / ask 9.5. The market is pricing roughly a 91 percent chance. You buy 100 YES at 90.9 cents, paying $90.90. Two outcomes:

• Event resolves YES: each contract pays $1, you collect $100, a $9.10 gross profit.
• Event resolves NO: contracts pay $0, you lose the $90.90 you staked.

You do not have to hold to resolution. If sentiment shifts and YES climbs to 95 cents, you can sell your position into the book and bank the difference, exactly like exiting any other tradable instrument.

Resolution and settlement

Resolution is the call on what actually happened. Settlement is the money moving from the losing side to the winning side. The two venues handle resolution differently, and the difference matters for risk.

• Kalshi resolves centrally. Each market lists named source agencies in its contract terms, filed with the CFTC. After resolution, settlement usually lands within a few hours.
• Polymarket resolves through the UMA optimistic oracle on Polygon. A proposer posts an outcome with a bond; if uncontested it finalizes after a challenge window of about two hours, but a dispute escalates to a UMA token-holder vote and can take several days. Oracle resolution carries its own tail risk. In March 2025 a UMA holder voting roughly a quarter of one dispute round (about 5 million UMA across three accounts) pushed a roughly $7 million Ukraine market to a contested 'Yes.' A larger episode followed in June and July 2025, when whales swung the resolution of a Zelenskyy-suit market that traded well over $100 million in volume. In August 2025, UMA moved Polymarket resolutions to a managed oracle (UMIP-189) that restricts who can submit proposals to a vetted whitelist.

Fees differ too, and they have been changing. Kalshi charges a per-trade taker fee of round_up(0.07 x C x P x (1 - P)) per contract, where C is contracts and P is price in dollars. That curve is parabolic: it peaks near 1.75 cents per contract at a 50-cent price and falls toward zero at the extremes, and there is no separate fee at settlement. Polymarket historically charged no trading fee, but starting in early 2026 it began applying taker fees in high-frequency crypto markets and select sports markets, while many event and world markets stay fee-free; on-chain trades also carry small Polygon gas. Check the current schedule for the specific market you are trading. Always price fees into your edge before you trade. On a contract that pays a 9-cent gross profit, a 2-cent round-trip cost is more than a fifth of your return.

From understanding the book to trading it

Once the mechanics click, the natural next step is testing whether your read on probability actually beats the market's. That is a quantitative question, and it rewards systematic discipline over gut calls. A rule like buy YES when implied probability sits 8 points below your model and the spread is under 2 cents is trivial to state and tedious to execute by hand across dozens of markets. That is where automation earns its keep.

Banger (bangertrades.com) is built for this. You write a Python strategy against the live Kalshi and Polymarket order books, paper-trade it to see how it would have filled at real depth, then run it live under a declarative risk envelope: per-trade cap, daily loss stop, max open positions, kill switch. You bring your own venue keys; Banger never custodies funds. US persons trade legally on Kalshi and Polymarket US, both CFTC-regulated; the international polymarket.com site blocks US persons. Whether you automate or click manually, the edge starts with the same two facts: read the book, and remember the price is a probability, not a guess.