Prediction marts are markets where people trade contracts whose payoff depends on the outcome of a future event. They are simultaneously a forecasting mechanism, a financial market, an information-aggregation system, and—depending on the contract and jurisdiction—a form of speculation that can resemble betting.
They have existed in various forms for decades, but they have become dramatically more visible with the bigger platforms. The CFTC describes the products traded on these markets as event contracts. (Commodity Futures Trading Commission)
Imagine a market asking:
Will the Federal Reserve cut interest rates at its September 2026 meeting?
A contract might trade at $0.65.
If the event happens:
YES contract → pays $1.00
NO contract → pays $0.00
If the event doesn't happen:
YES → $0
NO → $1
So, approximately:
$0.65 price = 65% implied probability
This is the fundamental idea behind a binary prediction market. The CFTC describes event contracts in essentially these terms: a contract can be priced between $0 and $1, with the price representing the market's collective implied probability, and the winning side receiving $1. (CFTC Public Comments)
The important distinction is that you're generally not simply asking people for their predictions.
Instead, you're allowing them to put money behind their beliefs.
Suppose the market is:
Will Candidate A win the election?
Trader Belief Action
Alice 70% Buys YES
Bob 40% Buys NO
Carol 65% Buys YES
David 30% Buys NO
Their trading creates a market price.
If the price settles around $0.62, the market is effectively saying:
"Given the information available to traders right now, we estimate roughly a 62% probability that Candidate A wins."
The price changes whenever traders change their views.
For a simple binary contract paying $1 if an event occurs and $0 otherwise:
Expected value = probability × $1
Therefore:
$0.20 → approximately 20%
$0.50 → approximately 50%
$0.75 → approximately 75%
$0.95 → approximately 95%
But there is an important caveat:
The market price is an implied probability, not a guaranteed or necessarily perfectly calibrated probability.
Prices can incorporate risk preferences, liquidity, fees, market structure, trading restrictions, arbitrage opportunities, and behavioral biases. Fidelity similarly emphasizes that prediction-market prices are market-based estimates rather than guarantees. (Fidelity)
Suppose you believe:
There is an 80% chance it will rain tomorrow.
The market is pricing YES at $0.60.
You think the market is underestimating the probability.
You buy YES at $0.60.
Your contract pays $1.
Your gross profit:
$1.00 − $0.60 = $0.40
That's a 66.7% return on the $0.60 purchase price, before fees.
The contract pays $0.
You lose:
$0.60
This illustrates something crucial:
Prediction-market trading is not simply about being right.
You need to determine whether your estimate of the probability is better than the market's estimate after considering the price and fees.
They can look very similar, but there are important conceptual differences.
You might see:
Team A +150
The sportsbook generally establishes odds and takes the other side of the customer flow, managing its exposure.
You might see:
Team A YES — $0.43
Participants trade against one another, and the market price emerges from supply and demand.
This makes prediction markets closer in structure to financial markets than to a conventional bookmaker, although the underlying subject can look very similar to gambling.
This distinction is one reason prediction markets have become such a significant regulatory issue.
Almost anything that can be expressed as a clearly measurable future outcome.
Who will win an election?
Which party will control Congress?
Will a bill pass?
Will a politician resign?
Will a government shut down?
Will inflation exceed 3%?
Will GDP growth exceed a particular level?
Will unemployment rise above a threshold?
Will the Fed change interest rates?
Will Bitcoin exceed $150,000?
Will the S&P 500 finish above a particular level?
Will a company exceed a particular valuation?
Will Singapore's temperature exceed 34°C?
Will a hurricane make landfall?
Will rainfall exceed a particular amount?
Who will win?
Will a team score more than X points?
Will a player achieve a particular statistic?
Will a movie exceed a particular box-office number?
Will a show win an award?
Will a company launch a product?
Will a CEO leave?
Will a merger close?
Will a ceasefire occur?
Will a government fall?
Will a particular diplomatic event happen?
This breadth is both one of prediction markets' greatest strengths and one of their biggest controversies.
The fundamental economic argument is information aggregation.
Imagine 100 people each know something different.
One follows economic data.
Another works in the industry.
Another understands polling.
Another tracks satellite imagery.
Another has expertise in monetary policy.
Another has studied historical election turnout.
Individually, each person has incomplete information.
A market gives them an incentive to express those beliefs financially.
The resulting price can potentially combine all of that information.
This is closely related to Friedrich Hayek's idea that markets can aggregate dispersed information.
Consider two questions.
"Who do you think will win?"
You can answer whatever you want.
There is little cost to being wrong.
"Would you put $1,000 behind your prediction?"
Now your incentives change.
If you are extremely confident, you may trade aggressively.
If you aren't confident, you may stay out.
That creates a mechanism for revealing the strength of beliefs, not merely their direction.
This is one reason prediction-market researchers have studied them as forecasting mechanisms for decades.
Prediction markets are a particularly interesting implementation of the wisdom of crowds.
The basic idea is:
A sufficiently diverse group can sometimes produce a better estimate than any individual member.
But the conditions matter.
A crowd doesn't automatically become wise.
A prediction market works best when:
participants have different information;
participants are financially incentivized;
participants can trade freely;
information is reasonably available;
prices can adjust rapidly;
there is sufficient liquidity;
manipulation is difficult;
traders are not all suffering from the same bias.
Research using the Iowa Electronic Markets has found substantial evidence that market prices can provide useful forecasts, although accuracy varies with market design and conditions. (Iowa Research Online)
This is one of their greatest advantages over conventional forecasting.
Suppose the market says:
Candidate A: 62%
Then breaking news arrives.
A major scandal is confirmed.
Within seconds or minutes, traders can reassess the information.
The price might move:
62% → 48% → 35%
Traditional polling cannot react nearly as quickly.
A prediction market therefore provides something like a real-time probability dashboard.
This distinction is extremely important.
A poll asks:
"Who are you planning to vote for?"
A prediction market asks:
"What probability do you assign to each candidate winning?"
These are fundamentally different questions.
A candidate can have:
48% support in polls
but only a 30% chance of winning
because the other candidate might have a more favorable electoral map.
Prediction markets attempt to estimate the probability of the eventual outcome, rather than simply measuring current opinion.
Research on the Iowa Electronic Markets found instances where market forecasts outperformed polls, particularly further ahead of elections. One study comparing IEM forecasts with 964 polls found the market closer to the eventual outcome 74% of the time. (ScienceDirect)
More recent research on Iowa's binary winner-take-all markets also finds evidence of reasonably well-calibrated probabilities. (Tippie College of Business)
This is a subtle but extremely important point.
Suppose a prediction market gives 100 different events a probability of 70%.
If the market is perfectly calibrated, approximately 70 of those events should actually occur.
It doesn't mean every individual 70% prediction will be correct.
A 70% prediction can absolutely be wrong.
That's not necessarily a forecasting failure.
The correct question is:
Across many predictions, do events priced at 70% happen roughly 70% of the time?
This is called calibration.
Suppose:
Market price = $0.60
You estimate:
Probability = 75%
Your expected payoff is:
0.75 × $1 = $0.75
You pay $0.60.
Your expected gross profit:
$0.75 − $0.60 = $0.15
That's a positive expected-value trade.
But suppose you estimate only a 55% probability.
Expected payoff:
0.55 × $1 = $0.55
Against a $0.60 purchase price, that's negative expected value.
So a serious prediction-market trader isn't simply asking:
"Will this happen?"
They're asking:
"Is the market probability wrong enough for me to have positive expected value?"
That's a much more sophisticated question.
Many prediction markets operate using an order book, similar to a stock exchange.
You might see:
YES
Best buyer: $0.62
Best seller: $0.64
The difference is the bid-ask spread.
If you immediately buy, you might pay $0.64.
If you immediately sell, you might receive $0.62.
The spread compensates liquidity providers and represents a transaction cost.
As liquidity increases, spreads can become narrower.
A prediction market needs people willing to provide liquidity.
A market maker continuously posts bids and offers.
For example:
Buy YES at $0.48
Sell YES at $0.52
If traders transact against those orders, the market maker earns the spread, while taking on risk that the market moves against them.
Professional market makers can make prediction markets substantially more liquid.
Prediction markets can also create arbitrage opportunities.
Suppose:
YES = $0.45
and
NO = $0.45
Together:
$0.90
If exactly one must win and each winning contract pays $1, buying both costs $0.90 and guarantees $1.
Gross arbitrage profit:
$0.10
This is an extreme simplified example, but arbitrageurs play an important role in keeping markets internally consistent.
Not everything is YES/NO.
For example:
Who will win the election?
You might have:
Candidate A — 45%
Candidate B — 35%
Candidate C — 15%
Other — 5%
The probabilities should approximately sum to:
100%
There can also be markets such as:
Will inflation be below 2%?
2–3%?
3–4%?
Above 4%?
These markets can create a probability distribution rather than a single binary probability.
Another form of prediction market asks participants to predict a quantity rather than a yes/no outcome.
For example:
What will Candidate A's vote share be?
Instead of:
YES = $0.65
you might have a forecast such as:
Candidate A expected vote share = 47.2%
The Iowa Electronic Markets historically used such vote-share contracts, which is important because some of the foundational research on prediction-market accuracy comes from these markets. (Cambridge University Press)
Market design matters enormously.
A good contract should have:
Bad:
"Will there be major political turmoil?"
Good:
"Will Country X's president resign before December 31, 2026?"
For example:
government statistics
official election results
a specified agency
a recognized sporting authority
Exactly when does the contract expire?
What happens if:
the event is delayed?
the organization changes its definition?
the result is disputed?
the data is revised?
A market with three traders isn't necessarily a useful crowd forecast.
Imagine a market asking:
"Will there be a recession in 2026?"
What exactly does "recession" mean?
Does it mean:
two consecutive quarters of negative GDP?
an official NBER recession?
a specific economic indicator?
a platform-defined threshold?
Different definitions produce different answers.
Therefore, prediction-market contracts need precise resolution rules.
The market isn't just predicting an event.
It is predicting an explicitly defined event under explicitly defined rules.
Prediction markets aren't magical.
They can suffer from:
One trader can move the price substantially.
A wealthy trader may intentionally move prices.
Traders may follow the crowd rather than independently evaluating information.
One participant may know something others don't.
Someone with privileged information may trade before the public knows.
Traders may bet according to what they want to happen rather than what they actually believe.
People may assign excessive confidence to their forecasts.
Poorly written resolution rules can create disputes.
Users ultimately depend on the exchange or settlement mechanism.
This is arguably one of the most important issues facing prediction markets.
Suppose a government employee knows something that hasn't been publicly announced.
They could potentially make a large trade based on that information.
The market price would then incorporate information that ordinary participants don't have.
This isn't merely an abstract concern.
The CFTC issued a 2026 advisory concerning enforcement cases involving misuse of nonpublic information and fraud in prediction markets. One case involved a political candidate trading on a contract concerning his own candidacy. (Commodity Futures Trading Commission)
More recently, researchers have raised allegations that some Polymarket wallets may have traded using nonpublic military information, although the research itself acknowledges that suspicious trading patterns don't prove insider trading. (Reuters)
Imagine a market:
Candidate A = 45%
A wealthy trader buys $10 million worth of YES contracts.
The price jumps:
45% → 60%
Now millions of people see:
Candidate A — 60%
Some may interpret that as new information.
They buy.
The original trader can potentially sell into the increased demand.
This is sometimes called information signaling or price manipulation, depending on intent and circumstances.
The key question is whether the trader is:
revealing genuine information
or
deliberately creating a misleading signal.
There's an interesting feedback loop:
Information → trader → trade → price → public information
Suppose a person discovers something important.
They trade.
The price moves.
Journalists notice.
Researchers notice.
Other traders investigate why the price moved.
They discover additional information.
They trade.
The market becomes even more informative.
In that sense, prediction markets can function as information discovery systems, not merely forecasting tools.
A common mistake is saying:
"The market says 80%, therefore the event has an 80% chance of happening."
A better interpretation is:
"Given the participants, information, incentives, liquidity and market structure at that moment, the market is pricing the event at roughly an 80% implied probability."
That distinction matters.
Markets can be wrong.
Stock markets can be wrong.
Bond markets can be wrong.
Prediction markets can be wrong.
The interesting question is whether they're wrong less often than alternative forecasting mechanisms.
There are two major ways of thinking about prediction markets.
The market exists primarily to produce information.
Example:
"What probability does the crowd assign to a recession?"
The contracts are tradable financial instruments.
Participants can:
speculate
hedge
arbitrage
provide liquidity
manage risk
The CFTC explicitly describes event contracts as derivatives that can be used for hedging or speculation. (Commodity Futures Trading Commission)
This second model is increasingly important as prediction markets become larger.
Suppose you're a business whose revenue is heavily affected by a hurricane.
A hurricane-related event contract could potentially allow you to offset some of that risk.
Or imagine a company whose revenues depend on a particular economic indicator.
A contract tied to that indicator could potentially serve as a hedge.
This is one reason regulators and financial-market proponents argue that event contracts shouldn't automatically be viewed as gambling.
The CFTC explicitly says event contracts can help businesses and individuals hedge event-driven risks. (Commodity Futures Trading Commission)
A statistical model might say:
Candidate A: 61.4%
A prediction market might say:
Candidate A: 58%
What's different?
A model relies on:
historical data
statistical assumptions
selected variables
model structure
A market incorporates:
human judgment
private information
interpretation of news
statistical models
expert knowledge
incentives
trading behavior
A prediction market is therefore an ensemble of models and human judgments, although its output can also be distorted by behavioral factors.
An expert panel might consist of:
20 economists
A prediction market could consist of:
20,000 participants
But bigger doesn't automatically mean better.
The market participants need to have:
relevant information
incentives to trade
capital at risk
sufficient independence
A small market of highly informed traders can potentially outperform a large crowd of uninformed traders.
Liquidity is one of the most important variables.
Suppose a market has only $10,000 of liquidity.
A $50,000 trade could move the market dramatically.
That doesn't necessarily mean the underlying probability changed dramatically.
It might simply mean:
there isn't enough opposing liquidity.
Therefore:
Price ≠ probability in isolation.
You also need to examine:
volume
open interest
bid/ask spread
depth
number of participants
recent trades
High volume generally makes a market more informative.
But volume alone doesn't guarantee quality.
A market can have enormous trading volume while being:
highly speculative
dominated by a small number of traders
driven by bots
subject to correlated beliefs
The academic evidence from Iowa suggests that characteristics such as market structure, contract type and trading volume can affect predictive performance. (Iowa Research Online)
There are several broad trading strategies.
Estimate the probability independently.
Then compare:
Your probability vs. market probability
Trade when you possess information that hasn't been fully incorporated into the price.
Exploit inconsistencies between related contracts.
Look for historical patterns in pricing.
Provide liquidity and capture bid/ask spreads.
Trade many contracts rather than making one large prediction.
A sophisticated trader might maintain a table like:
Event Market My probability Edge
A 45% 55% +10 pp
B 70% 72% +2 pp
C 20% 35% +15 pp
The trader isn't necessarily interested in the event with the highest probability.
They're interested in the largest discrepancy between market price and estimated probability, adjusted for:
uncertainty
liquidity
fees
correlation
time
capital requirements
Prediction markets also connect naturally to bankroll management.
If you believe an event has probability p, and the market offers attractive odds, you can theoretically determine an optimal fraction of capital to wager using the Kelly criterion.
The principle is:
Bet more when your information edge is larger, but never bet so much that one incorrect prediction destroys your bankroll.
In practice, traders often use fractional Kelly because probability estimates themselves are uncertain.
This is particularly important in prediction markets because even a 90% probability event fails 10% of the time.
Prediction markets are naturally Bayesian.
Start with:
Prior probability
Then receive new information.
Update:
Posterior probability
For example:
Initial:
Candidate A = 40%
New poll:
+5 percentage points
Major endorsement:
+7
Economic shock:
−10
Debate performance:
+4
New information continually changes the probability.
The market essentially performs a distributed, decentralized version of Bayesian updating.
Prediction markets force us to distinguish:
What people believe will happen
from
What people want to happen
and from
What people think other people will believe will happen.
That produces multiple layers.
A trader might think:
"Candidate A has a 40% chance."
But they might also think:
"The market believes 30%, and I'm going to buy because I think it will eventually move toward 40%."
They are therefore trading not only on the outcome, but sometimes on future market expectations.
That's very similar to financial markets.
The technology has improved substantially.
Modern platforms can combine:
real-time trading
mobile interfaces
automated market makers
APIs
blockchain settlement
stablecoins
global internet access
algorithmic trading
social-media distribution
This has made prediction markets dramatically easier to access.
Recent market growth has been enormous. A 2026 CFTC document cited roughly $12 billion in combined Kalshi and Polymarket trading in December 2025, around four times the comparable level a year earlier. (CFTC Public Comments)
Prediction markets are now moving from a niche forecasting experiment toward something closer to a new financial-market category.
The regulatory debate in the United States is especially significant.
The CFTC has argued that event contracts on CFTC-regulated exchanges fall within federal derivatives jurisdiction. (Commodity Futures Trading Commission)
At the same time, several states have challenged prediction markets under gambling laws, producing an increasingly significant federal-versus-state regulatory conflict. In April 2026, the CFTC sued Wisconsin over the state's actions involving several prediction-market companies. (Commodity Futures Trading Commission)
The issue remains highly active.
This is probably the central conceptual debate.
Critics say:
You're putting money on whether something happens. That's gambling.
Supporters say:
These are financial derivatives that allow risk transfer, price discovery and information aggregation.
Both perspectives have some validity.
The classification depends partly on:
contract design
jurisdiction
regulation
whether the participant is hedging or speculating
how the market is operated
That's why the regulatory treatment differs substantially across countries.
Political prediction markets are particularly controversial.
Potential benefits:
provide real-time probabilities
aggregate political information
reveal uncertainty
potentially improve journalism
provide an alternative to polls
Potential risks:
election manipulation
insider trading
misinformation
political gambling
market-driven narratives
attempts to influence voter behavior
Election officials have recently expressed concerns about these risks ahead of the 2026 U.S. midterms. (WIRED)
This may ultimately be the biggest opportunity.
Imagine every major question having a live probability:
Recession next year — 31%
Rate cut in September — 68%
Company X acquisition — 74%
Hurricane landfall — 22%
Candidate A wins — 57%
Product launch before December — 81%
Instead of consuming news as:
"Experts say X might happen."
you could see:
Current market probability: 63%
And then watch the probability change as new information arrives.
That potentially creates a new layer of the information economy.
This is a particularly interesting development.
Traditional media:
Event → journalist → article → reader
Prediction market:
Event → traders → price → reader
The price itself becomes a piece of information.
A journalist might write:
"Markets now assign a 72% probability to a rate cut."
The prediction market therefore becomes a data source for journalism.
There is also a potentially huge technology opportunity.
Instead of humans visiting a prediction-market website, applications could consume probabilities programmatically.
For example:
Fed cut probability: 68%
Recession probability: 31%
Candidate A win probability: 57%
Bitcoin > $150k probability: 42%
An AI system could then combine those probabilities with:
economic data
news
financial markets
polling
historical data
expert forecasts
This creates the possibility of probability-aware AI systems.
AI is particularly interesting because AI can participate in prediction markets in several ways.
An AI agent analyzes:
news
financial data
social media
historical statistics
and trades contracts.
AI explains:
"The probability moved from 42% to 61%. Here's why."
AI continuously provides bids and offers.
AI generates an independent probability that humans can compare with the market.
Humans provide:
intuition
domain expertise
unusual information
AI provides:
scale
consistency
rapid analysis
statistical modeling
The combination could be powerful.
If you are evaluating the industry seriously, I would focus on these:
Small markets can produce unreliable prices.
Rules differ across jurisdictions and are rapidly changing.
This can undermine the fairness of the market.
Large participants can sometimes move thin markets.
Poor resolution rules can create disputes.
The people participating aren't necessarily representative of the population.
Traders aren't perfectly rational.
Markets can underestimate rare events.
The exchange, oracle or settlement mechanism can fail.
Some events arguably shouldn't be financialized.
Requirements:
Clear question
Reliable resolution
Independent participants
Strong incentives
Liquidity
Transparent rules
Low manipulation
Good information flow
=
Potentially powerful forecasting mechanism
Remove several of those components and the market can become much less useful.
If you're studying prediction markets, these terms are worth knowing:
Term Meaning
Event contract Financial contract tied to an event
Binary contract Contract with two possible outcomes
YES Position that the event will occur
NO Position that it won't
Implied probability Probability suggested by the market price
Order book List of buy and sell orders
Bid Highest price someone will pay
Ask Lowest price someone will accept
Spread Difference between bid and ask
Liquidity Ability to trade without moving price substantially
Market maker Participant providing liquidity
Settlement Process determining the final payout
Resolution Determination of the actual outcome
Arbitrage Exploiting inconsistent prices
Calibration Whether probabilities match long-run frequencies
Information aggregation Combining dispersed information through trading
Event contract Common regulatory/financial term for prediction-market contracts
Prediction markets sit at the intersection of five industries:
Because contracts are bought and sold.
Because participants can win or lose money based on uncertain events.
Because prices can represent probabilities.
Because markets can aggregate huge amounts of information in real time.
Because market probabilities can become information consumed by the public.
That combination explains why they are attracting so much attention.
The most compelling argument for prediction markets isn't:
"People should be able to bet on everything."
It's:
Markets can provide a continuous, incentive-compatible mechanism for aggregating information about uncertain future events.
That is a much bigger idea.
The strongest use cases are likely to be situations where:
the outcome is objectively measurable;
information is dispersed;
new information arrives continuously;
conventional forecasting is slow or expensive;
participants have strong incentives to be accurate.
The weakest use cases are likely to involve:
extremely thin markets;
ambiguous outcomes;
heavy insider-information advantages;
highly manipulable events;
contracts where participants can influence the outcome they're trading.
And the biggest long-term question is whether prediction markets become primarily a new form of financial speculation or a global infrastructure for measuring probabilities.
They could ultimately become both.
In one sentence: a prediction market turns collective beliefs about the future into a continuously updated, tradable probability.
As of August 2026, the U.S. regulatory environment is particularly important to watch. The CFTC continues to assert federal jurisdiction over regulated event-contract markets, while states have challenged some prediction-market activity under gambling laws. At the same time, concerns about insider information and manipulation have become more prominent as trading volumes have grown. (Commodity Futures Trading Commission)