What is a prediction market?
The one-paragraph answer#
A prediction market is a market in contracts whose payoff depends on a future event. The standard contract, an event contract or binary contract, pays $1 if the event happens ("Will the Fed cut rates in December?") and $0 if it does not. Anyone who thinks the event is more likely than the current price implies can buy; anyone who thinks it less likely can sell (or buy the No side). The price at which they agree is a number between 0 and 1 that summarises what the participants, weighted by how much they are willing to put behind their views, believe.
If a Yes contract trades at 63¢, a buyer pays 63¢ to receive $1 if the event happens, a profit of 37¢, or loses the 63¢ if it does not. Ignoring costs, buying is a good trade only if you think the probability is above 63%. That is why the price can be read as a probability.
How a prediction market works#
Listing. The venue writes a contract: the question, the resolution source (an agency release, an official result, a named data provider), the time window and the rules for edge cases such as postponements, ties or revisions. These rules matter more than most participants realise; they define what the contract actually pays on.
Trading. Most large venues run a central limit order book: participants post bids and offers, and trades happen when they cross. Some markets instead use an automated market maker, an algorithm that always quotes and moves its price as it is traded against (see AMMs for prediction markets). Yes and No are two sides of one contract, so buying No at 37¢ is economically the same as selling Yes at 63¢.
Collateral. Every contract is fully collateralised. The buyer of Yes at 63¢ puts up 63¢, the buyer of No puts up 37¢, and the $1 is held until resolution. There is no leverage and no margin call; the most you can lose is what you paid.
Resolution and settlement. When the event is determined, the contract resolves to 1 or 0 under its rules, winning contracts pay $1 and collateral is released. Some venues determine outcomes themselves under their rulebook; on-chain venues may use an oracle with a dispute window.
| Step | What happens | What can go wrong |
|---|---|---|
| List | Venue writes question, source, window, edge-case rules | Ambiguous wording; a source that can be revised |
| Trade | Bids and offers meet in an order book or AMM | Thin liquidity; wide spreads |
| Hold | Collateral locked until resolution | Capital tied up for months |
| Resolve | Outcome determined from the named source | Disputes; edge cases; delays |
| Settle | Winners paid $1, losers $0 | Settlement timing differs by venue |
Why prices are probabilities, and when they are not#
For a risk-neutral trader with probability , the expected profit from buying Yes at is . In a market of such traders, the price settles where buyers and sellers balance, which under reasonable conditions is close to the average belief. Manski (2006) showed that with heterogeneous beliefs and budget constraints the price need not equal the mean belief, but it lies within a range around it; Wolfers and Zitzewitz (2006) show that under common utility assumptions the price is close to the mean belief. Either way it is a good summary statistic, not a guaranteed probability.
In practice five things move a price away from the probability of the event:
- Spread. You buy at the ask and sell at the bid; the mid is the better estimate.
- Fees. A fee per contract raises the break-even probability (see fee arithmetic).
- Time value. Money locked for a year has a cost: a 63% event a year away is worth about 60.5¢ at a 4% rate.
- Biases. Longshots tend to trade above their true probability (the favourite–longshot bias).
- Rules. The contract pays on its resolution source, which can differ from the event as commonly understood.
Our pricing primer works through each with numbers.
Types of prediction market#
| Type | Examples of the model | Money | Price formation |
|---|---|---|---|
| Regulated exchange | US designated contract markets listing event contracts | Real money, fiat | Order book with professional market makers |
| On-chain market | Blockchain-settled markets with stablecoin collateral | Real money, crypto | Order book or AMM; oracle resolution |
| Play-money / reputation | Forecasting platforms with points or leaderboards | None or virtual | AMM or aggregation of forecasts |
| Internal / corporate | Markets run inside firms on project deadlines or sales | Virtual or small stakes | Usually an LMSR |
| Academic | University-run research markets | Small real stakes | Order book |
The distinction that matters most for price quality is liquidity. A market with a 1¢ spread and real depth aggregates far more information than one with a 10¢ spread, whatever the venue type.
Who trades, and why#
- People with a view. Participants who believe they know something, from a better read of polling to a better weather model.
- Hedgers. A business exposed to an outcome (a policy decision, a weather event) can offset that exposure. This is the original economic argument for event contracts on regulated exchanges.
- Market makers. Professional firms, including ours, that quote both sides continuously and earn the spread while managing inventory. They supply the liquidity that makes the price meaningful; see market making in prediction markets.
- Relative-value traders. Participants who keep prices consistent across related contracts and venues, enforcing the no-arbitrage conditions.
- Entertainment. A large share of retail volume, especially in sport. It is often uninformed, which is part of what pays market makers to provide liquidity.
How accurate are prediction markets?#
Long-running evidence is favourable. Berg, Nelson and Rietz (2008) compared the Iowa Electronic Markets with 964 polls across the five US presidential elections from 1988 to 2004 and found the market closer to the eventual outcome 74% of the time, and significantly better than polls more than 100 days out. Wolfers and Zitzewitz (2004) survey similar results across domains. Rhode and Strumpf (2004) document that historical presidential betting markets in New York were remarkably accurate before polling existed.
Recent studies of exchange-traded contracts find good calibration overall, with systematic patterns: prices compressed toward 50% in some political markets, tail distortions, and accuracy that improves sharply as resolution approaches. The honest summary is that liquid prediction markets are among the best available forecasts, and that thin or low-attention markets should be read with care. Measuring this properly takes more data than most people expect; see how many forecasts it takes to evaluate a forecaster.
A worked example#
A contract asks whether a monthly inflation print will be at or above 3.0%. The order book shows 41¢ bid and 43¢ offered.
- The market's implied probability is about 42% (the mid).
- Buying 100 Yes contracts at 43¢ costs $43 plus a fee. At the 0.07 coefficient in Kalshi's published taker schedule, ≈ $1.72, so about $44.72 in total.
- If the print is 3.0% or higher, the contracts pay $100: profit $55.28. Otherwise the $44.72 is lost.
- The trade has positive expected value only if the true probability is above 44.7%, not 43%.
The event contract calculator does this arithmetic for any price, size and fee.
Questions#
What is a prediction market in simple terms?
A market where people buy and sell contracts that pay $1 if an event happens and nothing if it does not. The price, between 0 and $1, reflects how likely participants collectively think the event is.
How do prediction markets make money?
Venues typically charge trading fees, earn interest on collateral, or both. Market makers earn the bid-ask spread. Traders profit when they buy contracts priced below the event's true probability.
Are prediction market prices accurate probabilities?
They are good estimates, often better than polls or experts, but not exact. Spreads, fees, the time value of money, the favourite–longshot bias and contract wording all create small, predictable gaps.
What is the difference between a prediction market and a poll?
A poll asks people what they intend or believe; a prediction market asks them to put money behind a forecast of the outcome. Markets update continuously and weight views by conviction.
References#
- Wolfers, J. and Zitzewitz, E. (2004). Prediction markets. Journal of Economic Perspectives, 18(2), 107–126.
- Wolfers, J. and Zitzewitz, E. (2006). Interpreting prediction market prices as probabilities. NBER Working Paper 12200.
- Manski, C. F. (2006). Interpreting the predictions of prediction markets. Economics Letters, 91(3), 425–429.
- Berg, J. E., Nelson, F. D. and Rietz, T. A. (2008). Prediction market accuracy in the long run. International Journal of Forecasting, 24(2), 285–300.
- Rhode, P. W. and Strumpf, K. S. (2004). Historical presidential betting markets. Journal of Economic Perspectives, 18(2), 127–141.
DisclosureThis note is research and education. It is not investment advice or an offer to trade, and nothing in it describes the firm's own results.