What is a prediction market?

If you've watched the news lately, you've seen a new kind of number. Not a poll, not a pundit's hunch — a hard percentage that updates live: this candidate is 68% likely, that rate cut is 40% likely. As of December, both CNN and CNBC started putting these odds on screen.¹ They come from something called a prediction market.

This is the whole thing, start to finish — what a prediction market is, how the mechanism actually works under the hood, and why it may be one of the more important financial ideas of the decade. We'll start with no jargon and no math heavier than a coin, then go as deep as it's worth going. Everything builds on one idea, so let's get it right.

The world's simplest contract

A prediction market is a place to buy and sell one very simple thing: a contract that pays you $1 if some event happens, and nothing if it doesn't.

That's the whole object. Pick an event — say, "Will it rain in Hanoi tomorrow?" Someone writes a contract: it pays $1 if it rains, $0 if it stays dry. Now people can buy and sell that contract, the same way they'd trade a share of stock.

If you were certain it would rain, you'd happily pay almost a full dollar for something about to pay you a dollar. If you were certain it wouldn't, you'd pay close to nothing. But the truth is usually somewhere in the middle — and that "in the middle" is where the magic is.

The price is the chance

Here is the one idea that makes the whole thing work. Because the contract pays exactly $1 when the event happens, the price people will pay tells you how likely they think it is.

Suppose the rain contract is trading at 62¢. Why 62, and not 90 or 10? Because the crowd — with real money on the line — collectively believes there's about a 62% chance of rain. The price and the probability are the same number wearing different clothes:

Read the price, and you've read the odds — no translation needed. And the number is alive: if the forecast worsens, buyers push the price up toward 100¢; if the skies clear, it slides toward zero. It moves with the world, in real time.

Why the number is worth trusting

A poll costs nothing to answer, and nothing to answer wrong. A market is the opposite: every opinion is a position, and being wrong costs real money. That single difference — skin in the game — is what makes the price sharp.

Picture the rain contract sitting at 50¢ when you've checked the radar and believe the true chance is 67%. You have a 17¢ edge on every share, so you buy — and your buying pushes the price up toward your estimate. The edge any trader sees is just the gap between what they know and what the price says:

Mispricings are money lying on the floor. The better-informed you are, the more you bet, and the harder you pull the price toward the truth. Money doesn't just measure confidence — it weights it, handing the loudest voice to whoever is most willing to be wrong in public for cash. In the 2024 U.S. election, one trader reportedly ran his own private polls — paying for better information than the public had — and moved size on the result, making tens of millions when he turned out right.² A poll could never reward that work. A market pays for it.

The effect shows up in the shape of the signal. A poll arrives every few days, lurches with sampling noise, and lags the news. A market is a line that moves the second something happens — because the first trader to understand the news has a profit waiting if they act before everyone else.

Market · continuous Polls · periodic, lagging

This isn't only intuition. Decades of research on market-based forecasting find that, when they're liquid enough to function, prediction markets beat polls, expert panels, and statistical models on a wide range of questions.³ The reason is structural, not magical — and it's the next idea.

It aggregates what no one knows alone

In 1945, Friedrich Hayek pointed out that the knowledge a society needs to make good decisions is never held in one place. It's scattered — "dispersed bits of incomplete and frequently contradictory knowledge" spread across millions of people, none of whom sees the whole picture.⁴ A price, he argued, is the device that pulls all those private fragments into one public number, so people can act on what others know without ever being told it directly.

A prediction market is that idea aimed straight at the future. Thousands of people each hold a sliver of relevant information — a poll a campaign hasn't released, a supplier's quiet warning, a doctor's read on a trial. None of them can forecast the outcome alone. The market lets each one stake their sliver, and the price collapses all of it into a single, living probability.

Because anyone can create a market, this sensor points at questions no exchange ever bothered with. Will this AI model top the benchmark next quarter? Will this scientific result replicate? Will the new airport open by June? Firms already run internal markets to forecast their own launch dates; researchers have used them to predict which experiments will hold up.⁵ Anyone can stand up a market for a question worth answering — and the crowd will price it. This is far older than the apps, too: as far back as the 16th century, bettors in Rome were quoting live odds on papal succession,⁶ and election betting was front-page news decades before scientific polling existed.

So who actually trades one?

Three kinds of people show up to a prediction market — and the third is the one that matters most for everyone else:

That last column is the quiet wonder of the whole design. Nobody sets out to produce a public forecast. They're all chasing their own profit. But the price they fight over becomes one — a number the rest of us get to read for nothing.

A forecast, not a bet

That second column — the hedge — is also where most people get prediction markets wrong, so it's worth being precise. A casino and a prediction market both involve money and uncertainty. They are opposites in the one way that counts.

At a casino, the house sets the odds and builds in a margin, so the average player loses by design and no information is produced — the roulette wheel knows nothing about the world. In a prediction market, nobody sets the odds; the price floats to the crowd's best estimate, the most accurate forecasters win money from the least, and the by-product is something valuable to everyone, player or not: a public, real-time probability. Your opponent isn't the house — it's the rest of the crowd.

The casino

House odds · fixed margin

• The operator sets the odds
• Luck-driven — no information value
• The house wins on average, by design
• A wager that simply wins or loses

The prediction market

Yes 62¢ · No 38¢ · the crowd's price

• The market discovers the price
• Information-driven — the better forecast wins
• The most accurate traders win, not the house
• A position you can buy or sell anytime

That distinction is what makes an event contract a financial instrument rather than a game — and it unlocks the most underrated use of all: hedging. If your business depends on a central-bank decision, a port opening on schedule, or a mild winter, you can take the other side of that risk and offload it. A prediction is something you watch. A hedge is something you own. Event contracts let you insure outcomes that no insurer will write a policy for.

Under the hood: who takes the other side?

So far we've described the price as if it sets itself. Time to open the hood. Recall the one equation everything rests on — for a contract that pays $1 on YES and $0 on NO, a risk-neutral trader buys below their estimate of the true probability and sells above it, so in equilibrium the price is pinned to the crowd's collective estimate:

That works beautifully once a market is busy. The problem is the first day, when it isn't. Early markets are thin: you want to bet, but there's no one to bet against. The fix is an automated market maker — a formula that will always quote a price. The classic is Robin Hanson's Logarithmic Market Scoring Rule (LMSR), which prices a whole basket of outcomes from a single cost function:

Here \(q_i\) is the number of shares outstanding on outcome \(i\), and \(b\) is a liquidity parameter — bigger \(b\) means deeper books and smaller price moves per trade. The cost to move the market from \(\mathbf{q}\) to \(\mathbf{q}'\) is simply \(C(\mathbf{q}') - C(\mathbf{q})\). Differentiate, and the instantaneous price of each outcome falls out as a softmax:

Two things to notice. The prices always sum to one — \(\sum_i p_i = 1\) — so the market is a coherent probability distribution by construction. And the maximum the market maker can ever lose is bounded by \(b \ln n\) for \(n\) outcomes: liquidity has a known, finite price. That boundedness is what makes a market operable rather than a charity — the operator knows the worst case before opening the doors.

What's still hard

None of this means the machine runs itself. Three problems stand between a clever idea and a market you can trust.

Liquidity. A new market is empty — you want to trade, but there's no one to trade against. The fix is the automated market maker above: a formula that always quotes a price, with a known, finite cost to the operator. That part is solved on paper; the LMSR has been the canonical answer for two decades.

Resolution. Someone has to declare what actually happened, cleanly and without dispute. If the outcome can be gamed or argued, the price means nothing. The answer is disciplined settlement — multiple independent sources, an auditable trail, fully collateralised payouts so nothing defaults.

Trust. A market is only a sensor if its price is honest, which means defending against manipulation and insider trading. The reassuring part is that a market fights back on its own: push the price away from the truth and you hand every other trader a profit for shoving it back. The unreassuring part is that self-correction isn't enough at scale — you need real surveillance, position limits, and someone accountable for the integrity of the book.

That last point is the one we spend our days on at Seeker. Two of these three problems are engineering. The third — trust — is mostly regulatory: in most countries a market that pays out on real-world events is a licensed financial product, and the right to run one is granted one jurisdiction at a time. That's the part people underrate. The mechanism is universal and copyable; the license is neither.

The money moving in is buying infrastructure

For most of its life this was a fascinating idea with no liquidity. For a decade prediction markets were an academic toy. Then the volume arrived: monthly trading went from under $100M in early 2024 to more than $13B by the end of 2025 — the bulk of it sports — and the sell side now models the category on a path to $1 trillion a year by 2030.⁸

Then the institutions arrived — and what they bought says what they think it is. The Intercontinental Exchange, which owns the New York Stock Exchange, committed up to $2B to Polymarket — structured not as a wager but as a deal to become the global distributor of its event data, piping crowd-implied probabilities to institutions alongside its securities feeds.⁹ Kalshi, the first federally-regulated exchange, was just valued at $11B; CME and Cboe are building their own event-contract products. This is infrastructure money, not novelty money — and it is buying the rails, not a seat at the table.

ICE → Polymarket

up to $2B

For the data, not the bets — ICE becomes the global distributor of Polymarket's event feed. Oct 2025.

Kalshi

~$11B

The first U.S.-regulated exchange, valued in its latest round. Dec 2025.

CME · Cboe

entering

The established derivatives exchanges are building their own event-contract products.

The one thing to remember

A prediction market is a place to trade simple contracts whose price is a live probability. Buy if you think the crowd is too low, sell if it's too high, and the price you all settle on becomes a forecast the whole world can read. The mechanism is well understood; the math is two decades old; the money is here. What isn't settled is trust.

That's the part we spend our days on at Seeker. It's the pattern crypto exchanges walked a decade ago — the U.S. legitimized the asset class, then every market opened it one regulator at a time, and the first licensed venue took most of the liquidity. Our bet is that prediction markets run that same playbook, and that the venue people trust in each market is the one that showed up early, built the surveillance and settlement in from the first trade, and won the license to operate inside the rules rather than around them.¹¹

The deeper reason to care has nothing to do with any one company. A society that can see its own best guess about the future — priced in the open, updated in real time, with money keeping everyone honest — makes better decisions than one arguing over stale polls and confident pundits. If we get the trust problems right, prediction markets stop being a place to gamble on the news and become part of the infrastructure we use to navigate what's coming. That's why they matter. But it all rests on the single sentence we started with — a price is a probability you can trade.