What is a prediction market?
A prediction market is a marketplace where people trade contracts whose payout depends on the outcome of a future event. Will inflation be above 3% next quarter? Will candidate X win the election? Will Bitcoin trade above $200,000 by year-end? Every prediction market reduces a messy question about the future to a number between 0 and 1, and lets anyone with capital trade against that number.
Each potential outcome trades as its own asset. The price of an outcome is interpreted as the market-implied probability of that outcome happening — a 67¢ contract that pays $1.00 implies the market thinks the event will occur 67% of the time. When the event resolves, holders of the winning side receive $1.00 per share; everyone else gets zero.
Prediction markets are a way to price uncertainty. The crowd's wallets become a real-time forecasting machine — one that updates the moment new information appears, not once a week when the next poll lands.
That last point is the difference that makes prediction markets uniquely useful. Polls are snapshots. Pundits are stories. A market price is a live, capital-weighted consensus that incorporates every news headline, every leaked memo, every backroom rumor, every analyst note — the moment any of those moves a participant to put real money behind a new estimate.
Why prediction markets work better than polls
Three forces make prediction markets surprisingly accurate forecasters, and each one solves a problem that polls and pundits famously suffer from.
Skin in the game
Participants must risk capital to trade, which filters out the cheap talk that fills up cable news. There is no upside to lying to a market — you only lose money. Survey responses, by contrast, are free, anonymous, and easy to spam. This is why a small group of well-informed traders can produce probabilities that systematically beat large but uninformed populations, exactly the result Wolfers and Zitzewitz documented in their canonical Prediction Markets review.
Information aggregation
Every news headline, leaked report, expert opinion, and private signal eventually shows up in price. James Surowiecki's Wisdom of Crowds and Francis Galton's famous ox-weighing experiment both describe the same phenomenon: large groups of independent estimators, when averaged, beat almost any individual estimate. Prediction markets are crowd wisdom made continuous and weighted by stakes.
Arbitrage discipline
Anyone with a different probability estimate can step in and trade the spread. If a market prices a 70% event at 50¢, sharp traders quickly buy until the price reflects reality. This is the same mechanism that keeps Treasury yields aligned across maturities — it just operates on probabilities of events instead of probabilities of cash flows.
Empirical research backs this up. The Iowa Electronic Markets, Kalshi, Polymarket, and Manifold have repeatedly outperformed polls and pundits at typical horizons, especially in the final weeks before resolution when calibration matters most.
A brief history of prediction markets
The idea is older than most people think. Wagering on political and economic outcomes dates back at least to the Wall Street betting markets of the 1880s and 1920s, where election odds were a daily fixture of the financial press. Modern prediction markets — designed to extract information, not just entertain bettors — emerged from three converging streams.
| Era | Milestone | Why it matters |
|---|---|---|
| 1988 | Iowa Electronic Markets launches | First academic prediction market under CFTC no-action relief; established the empirical case for market accuracy. |
| 2003 | Robin Hanson publishes LMSR | Automated market maker that solves the bootstrapping problem; no order book required. |
| 2014–2020 | PredictIt, Augur, Polymarket | Bring prediction markets to retail through regulated, blockchain, and offshore liquidity rails. |
| 2024+ | Kalshi, Polymarket scale | Regulated event contracts in the US and global liquidity pushing >$1B in monthly notional. |
Each step removed a constraint: bootstrap liquidity, lower the cost of market creation, and reach a global audience. The result is a category that finally looks tradable to institutional desks, not just hobbyists.
How prices form: order books vs automated market makers
Modern prediction markets use one of two pricing mechanisms.
Order book markets match buyers and sellers directly. Kalshi and the Iowa Electronic Markets both use this approach. Order books work when liquidity is abundant — at the busiest prediction markets, the bid/ask spread on the headline contracts can be as tight as a few basis points.
Automated market makers (AMMs) quote prices algorithmically, acting as the counterparty to every trade. They dominate decentralised platforms (Polymarket, early Augur) and most long-tail markets where there are not enough traders to fill an order book. Two AMM families matter:
- LMSR (Logarithmic Market Scoring Rule) — Robin Hanson's original design, still the cleanest implementation of bounded-loss market making.
- LSAMM — a modernised variant that adapts liquidity dynamically and is the algorithm behind several production platforms. We unpack the math in our dedicated post on the LSAMM algorithm.
Both AMMs guarantee that quoted probabilities sum to 1 across mutually exclusive outcomes, which is the property that lets you read prices as probabilities without doing extra math.
If you want a deeper comparison of the trade-offs between pricing engines, see comparing AMM liquidity models.
What prediction markets are not
There are common misconceptions worth addressing up front, because they shape how regulators, journalists, and new users approach the category.
- They are not gambling. Prediction markets resolve based on real-world data with public sources of truth. The closest analogue is futures trading on a structured event, which the CFTC has recognised as a legitimate use case.
- They are not opinion polls. Polls collect intent; markets collect bets. A poll asking "Who will you vote for?" answers a different question than "Where would you put $100 of your own money?"
- They are not always efficient. Liquidity, regulation, and emotion all create exploitable inefficiencies. The largest, most liquid markets approach efficiency. The long tail does not.
- They are not new. Wall Street had thriving political betting markets a century before Polymarket existed.
- They are not a black box. Every quoted price has a math-grounded interpretation: it is the probability at which the market is indifferent between Yes and No.
How to read a market: implied probability and edge
Reading a prediction market is mostly a matter of mapping the price to a probability.
If the Yes contract trades at 0.62, the market estimates a 62% chance the event occurs. The No contract trades at 0.38 (because mutually exclusive outcomes must sum to 1, ignoring fees). Your edge is the difference between your private estimate and the market price.
edge = p_you - p_market
If you believe the true probability is 0.70, your edge on Yes is +0.08 (8 percentage points). Whether that edge is large enough to trade depends on your spread, your position sizing rules, and the time until resolution — topics we cover in detail in risk management for prediction market traders and the bankroll framework playbook.
A simple read-the-tape checklist:
- Identify the resolution criteria. Who decides the outcome, and from what data?
- Note the last price and the spread. Is the bid/ask tight enough that your edge survives a round trip?
- Look at volume and open interest. Thin markets are often mispriced for a reason.
- Map the price to a probability you actually believe. If the gap is meaningful, you have a trade.
When prediction markets shine — and when they don't
Prediction markets are a tool, not a religion. They beat the alternatives in some scenarios and underperform in others.
Where they win
- Objectively resolvable questions with a clear data source (e.g., "Will Brazil's GDP growth exceed 2% in 2027 per IBGE?").
- Multiple stakeholders with dispersed private information — elections, sports, corporate earnings, monetary policy.
- Continuous probability updates, where a weekly poll is too slow and a market can move every minute.
- Cross-domain forecasting where no single model dominates and aggregation beats specialisation.
Where they struggle
- Tomorrow's weather — a meteorological model is faster and just as accurate.
- Hyper-local outcomes with no audience and no liquidity.
- Self-referential questions ("Will this market resolve Yes?") where reflexivity kills information value.
- Markets where the resolution criteria are ambiguous — the LIBOR settlement disputes of the 2010s are a cautionary tale.
The skill of running prediction markets at scale is building the markets that win and avoiding the ones that struggle. We explore that selection problem in category design for market discovery and AI-assisted market creation.
The economics of being a market participant
Trading prediction markets is not the same as trading equities or crypto. The payoff is bounded — you cannot lose more than the cost of your shares, and you cannot win more than $1.00 per share. That changes the optimal strategy in two ways.
Convex bets are different. A 5¢ Yes contract has a 19× upside. A 95¢ Yes contract has a 5% upside. Tail bets need to be sized accordingly, and the optimal Kelly fraction is much smaller than naïve intuition suggests. See our forecast calibration playbook for the math.
Information has a half-life. Most edges in prediction markets evaporate within hours of new information becoming public. The traders who win consistently are not the ones with the best secret models — they are the ones who execute fastest when their pre-built model meets a market that has not yet repriced.
Liquidity is uneven. Headline markets (US presidential election, Fed meetings) are deep enough to absorb seven-figure trades with minimal slippage. Long-tail markets can move 5–10 cents on a $500 order. Knowing where you sit on that liquidity curve is half the battle, which is why we wrote about market microstructure for forecasters.
Getting started as a builder or trader
For a builder mindset, start with three resources:
- Read Robin Hanson's original LMSR paper. It is short, well-written, and explains why market makers can be subsidised with bounded loss.
- Inspect a live order book on Polymarket or Kalshi and look for arbitrage across correlated markets.
- Run a small paper-trade book for two weeks. Track every trade you would have made, then compare your hit rate against the closing market price.
For a trader mindset, the path is different:
- Pick three markets you understand in depth — politics, sports, or macroeconomics — and ignore everything else for the first 90 days.
- Build a simple probability spreadsheet with your estimate, the market price, the spread, and the implied Kelly fraction.
- Track every trade in a journal that includes the thesis, exit plan, and outcome. Most edge in prediction markets comes from process discipline, not insight.
Prediction markets are an unusually transparent corner of finance. Every price is a probability, every trade is a vote, and every resolution is a public record. That is what makes them such a powerful forecasting tool — and why they are quietly becoming infrastructure for journalism, policy, and even corporate planning. For the broader picture, see our outlook on the future of prediction markets in 2026.
Frequently Asked Questions
Are prediction markets legal?
Legality depends on jurisdiction. In the United States, regulated venues like Kalshi operate as CFTC-designated contract markets, while platforms like Polymarket are restricted for US persons. In the EU, regulators treat most event contracts as derivatives, requiring authorisation under MiFID II. Many other countries fall into a grey zone where prediction markets are tolerated but not formally licensed. We unpack the global landscape in regulatory paths for prediction markets.
How accurate are prediction markets compared to polls?
Empirical studies — including the foundational work by Wolfers and Zitzewitz — show that prediction markets match or outperform polls in 60–80% of election cycles, especially in the final two weeks before resolution. The advantage shrinks for markets with thin liquidity or ambiguous resolution criteria, and disappears entirely when the underlying question is dominated by domain experts who do not trade.
What is the difference between a prediction market and sports betting?
Sports betting books price events to balance their own exposure and earn a vig (commission). Prediction markets price events to aggregate information and let users trade against each other. The same event can have very different prices on a betting site and a prediction market because the sportsbook is optimising for risk management while the prediction market is optimising for information discovery.
Can I lose more than I put in?
No. Prediction market contracts have bounded payoffs — you can lose at most the price you paid for your shares, and you cannot win more than $1.00 per share. There is no margin call, no liquidation, and no synthetic leverage on the headline contracts. This makes them structurally safer than futures or perpetual swaps for retail participants.
How does Polymarket make money?
Polymarket charges no trading fees on the underlying outcomes; instead, the platform earns through MEV-style spread capture and integrations with on-chain liquidity. Other platforms — Kalshi, PredictIt, regulated venues — charge a percentage of profits on winning contracts. The fee structure matters because it directly reduces the edge available to traders.
What are binary versus multiple-outcome markets?
A binary market has exactly two outcomes (Yes / No), while a multiple-outcome market has three or more mutually exclusive outcomes. Binary markets concentrate liquidity and trade tighter; multiple-outcome markets reveal a richer probability distribution but spread liquidity thin. We compare the trade-offs in binary vs multiple outcome markets.
Is there a beginner platform to practise without risking money?
Yes. Manifold Markets lets users trade with play-money "mana", which mirrors real prediction-market dynamics without financial risk. It is the best on-ramp for builders who want to feel an AMM in their hands before deploying capital on a regulated venue. The University of Iowa also maintains the IEM with small position limits for academic research.
How do prediction markets resolve disputes?
Each platform defines a resolution source — often a primary news outlet, an official data series (like the BLS for inflation), or a designated oracle. Disputes are resolved either by a centralised oracle committee (Kalshi), an on-chain dispute mechanism (Augur, UMA's Optimistic Oracle), or a hybrid model. Builders should always publish the resolution source before launch and avoid markets where the truth conditions are subjective.
Where to go next
You now have the conceptual scaffolding. The next steps depend on what you want to build:
- If you want to trade, start with risk management for traders and bankroll frameworks for information traders.
- If you want to understand the math, jump into the LSAMM algorithm and our deep dive on market microstructure.
- If you want to design markets, read category design and binary vs multiple outcome markets.
- If you want the macro picture, our future of prediction markets in 2026 outlook lays out where the industry is heading.
Welcome to the world of decoded probabilities.
