Venue Types

The same question can trade in several marketplaces at once. A contract on the outcome of an election, a rate decision, or a championship may trade on a federally regulated exchange, on a crypto-native venue, inside a brokerage app, and on a play-money platform simultaneously, each quoting its own price. What separates these venues is not the events they list but how they are built: who operates the market, where the money sits, what a position physically is, and who declares the outcome. Those structural differences hold steady even as individual platforms launch, merge, and rebrand, which makes them the useful level at which to compare. This guide maps the venue types and the mechanics behind each one. For the instrument itself, see Event Contracts; for how prices read as probabilities, see Prices and Probabilities.

What types of prediction market venues exist?

Most venues follow one of six structural models. The table below summarizes how each type handles the questions that matter most: what you deposit, how positions settle, who decides outcomes, and what stands between you and the market.

Four of these are marketplaces in the full sense: regulated exchanges, crypto-native venues, betting exchanges, and play-money platforms all run two-sided markets where participants trade against each other and prices emerge from that trading. Brokerage-embedded access is different in kind, because a brokerage is a doorway to an exchange rather than a venue of its own. Forecasting platforms sit at the far end of the spectrum, with no trading at all and probability estimates that come from scored forecasts rather than prices.

Betting exchanges are among the oldest of these structures. Platforms in the Betfair lineage have run peer-to-peer markets on sports and politics for more than two decades, with participants who "back" an outcome (comparable to buying Yes) or "lay" it (comparable to buying No) at prices quoted in odds formats rather than cents; Betfair's exchange guides document the mechanics. Backing an outcome at decimal odds of 4.0 puts 25¢ at risk on a position that settles at $1 if the outcome occurs, which makes it economically similar to buying a Yes contract at 25¢. The market mechanics are nearly identical; the main differences are vocabulary and regulatory category, since betting exchanges operate as licensed gambling products in their home jurisdictions rather than as regulated derivatives markets. The rest of this guide focuses on the other five types. For the contrast with fixed-odds sportsbooks, see Markets vs. Sportsbooks, and for how these structures evolved, see History of Prediction Markets.

How do regulated prediction market exchanges work?

In the United States, event contracts are regulated as derivatives, and the venues that list them are designated contract markets, the same legal category as commodity futures exchanges. Oversight sits with the Commodity Futures Trading Commission, whose explainer, Understanding Prediction Markets and Event Contracts, describes the model: fixed-settlement yes/no contracts listed by a registered exchange, with the customer protections that apply to derivatives markets generally.

Structurally, a regulated exchange runs a central limit order book. Participants post orders to buy Yes or buy No at prices between 1¢ and 99¢, and a matching engine crosses them. The exchange operates the marketplace and charges fees for trading; it does not take the other side of trades, a point venue documentation tends to address directly because new traders often assume otherwise (How does Kalshi make money?).

The mechanism that makes binary markets work is full collateralization. Suppose a contract trades at 34¢. An order to buy Yes at 34¢ can only execute against an order that effectively pays 66¢ for No. Together, the two sides deposit the entire settlement value:

$$\text{Yes price} + \text{No price} = \$1$$

In plain terms, every matched trade funds the full dollar that the winning side will receive at settlement, before fees. ForecastEx describes this pairing explicitly: incoming Yes and No orders match only when their prices combine to cover settlement. The direct consequence for traders is that the maximum loss on any position is the price paid for it, and nothing can be margin-called, because nothing is borrowed.

Settlement runs through central clearing. When a contract's outcome is determined, the clearinghouse credits holders of the winning side $1 per contract, and the losing side expires at zero. The source of truth for each contract, along with timing rules such as the difference between a market's close and its determination, is written into the contract terms, and the exchange settles according to those terms; Kalshi's market FAQs walk through the settlement questions traders ask most often. Deposits and proceeds are dollars throughout, identity verification is standard, and disagreements about a resolution go through the exchange's rules and, beyond that, the regulator's channels.

Reading the order book itself (the bids, asks, spread, and depth behind the headline price) is its own skill, covered in Order Books.

How do crypto-native prediction markets work?

Crypto-native venues implement the same economics with different plumbing. The clearest entry point is custody. Instead of dollars in an account at a regulated intermediary, funds are held as stablecoins, tokens designed to track the dollar, in a blockchain wallet. Depending on the venue, that wallet may be one you control directly or one created for you at signup. Either way, a position is not an entry in a clearinghouse ledger; it is a token you hold.

The $1 identity carries over exactly. On Polymarket's order book, a limit order to buy Yes at 60¢ matches a limit order to buy No at 40¢, and the combined $1.00 is converted into one Yes token and one No token, delivered to their respective buyers. Trading a share means transferring that token, and holding the winning token at resolution redeems it for $1 of collateral. This mint-and-redeem cycle is the on-chain equivalent of central clearing.

Matching architectures vary. Several large venues run hybrid order books, where an operator matches orders off-chain for speed while matched trades settle on-chain through smart contracts, a design Polymarket documents as combining centralized matching with non-custodial settlement. Others use automated market makers, where a liquidity pool quotes prices from a formula instead of a book of resting orders; the design descends from Robin Hanson's logarithmic market scoring rule, a mechanism built to keep prediction markets quoting prices even when few traders are present. The trade-offs between books and pools are part of the Order Books guide.

Resolution is where crypto-native venues diverge most from their regulated counterparts. There is no exchange rulebook backed by a regulator. Outcomes are instead determined by an oracle, a process for bringing real-world facts on-chain. In the optimistic-oracle pattern documented by UMA, anyone can propose an outcome by posting a bond, a challenge window follows, and if the proposal is disputed, the question escalates to a vote of the oracle's token holders. The bond makes false proposals costly, and the dispute path stands in for an appeals process. This design removes the need for a central referee, at the price of making resolution itself a process with participants and incentives.

As on regulated venues, the platform is not the counterparty. Traders trade against each other, and venue documentation tends to answer the question in as many words (Is Polymarket the house?). Access rules differ by jurisdiction and change over time, so the venue's own documentation is the reliable reference on who can trade where.

How does brokerage access to prediction markets work?

Retail participation often arrives through a brokerage rather than through a venue directly. The pattern is consistent from one operator to the next. Event contracts appear inside a brokerage app, alongside stocks and options, but the brokerage is not the marketplace. Orders route to a partner exchange where the contract actually lists, and positions clear there. Robinhood's event-contract documentation, for example, names the regulated exchanges its customers' orders route to. Some operators go a step further and acquire or build an exchange of their own to route into, which changes the corporate structure without changing the mechanics.

Both paths end at the same order book. What changes for the trader is the intermediary layer: the account, the interface, the commission schedule, and the product catalog belong to the broker, while the contract, the counterparties, and the settlement mechanics belong to the exchange. Brokerage catalogs are often a curated subset of what the underlying exchange lists, and the available order types may be simplified relative to trading on the exchange directly.

The brokerage route matters for two reasons. The first is distribution, because placing event contracts in front of people who already hold a funded account is often how a mainstream audience first encounters the category. The second is the audience itself. Brokerage education tends to frame event contracts as portfolio instruments as much as speculative ones, including using forecast contracts to hedge economic and climate risks, an angle retail-first venues rarely emphasize. Sportsbook operators that enter the category tend to follow the same structural playbook, fronting an exchange with a consumer brand. Whatever the brand on the app, the useful questions stay the same: who operates the exchange, who clears the trades, and who holds the funds.

What are play-money and forecasting platforms?

A separate family of platforms runs on stakes other than cash, and it plays a larger role in the ecosystem than its size suggests, because it is where much of the forecasting craft gets taught and measured.

Play-money markets work mechanically like real-money venues: two-sided trading, prices between 0 and 1, positions that settle when the question resolves. The difference is the currency. Manifold runs markets in a platform currency that cannot be converted to cash and lets any user create a market on any question. That combination produces enormous breadth. Questions too niche, too informal, or too hard to specify for a regulated listing process can still get a market and a price. Resolution typically falls to the market's creator or to platform processes rather than to an exchange rulebook or an oracle, which works at play-money stakes and would be hard to sustain at real-money ones.

Forecasting platforms drop trading entirely. On Metaculus, participants submit probability forecasts directly, the platform aggregates them, and accuracy is measured with scoring rules that reward well-calibrated forecasters over lucky ones. There is no order book and no price, only a running aggregate of the community's forecasts. Tournament platforms such as Good Judgment Open run the same idea as public competitions scored on the Brier scale, continuing the research lineage described in Superforecasting.

The oldest bridge between these worlds is academic. The Iowa Electronic Markets, a University of Iowa research and teaching project, began running real-money election markets at small position limits in 1988 and supplied much of the early evidence that market prices forecast well. The longer arc from academic experiments to today's venues is traced in History of Prediction Markets.

For a trader on real-money venues, these platforms serve three purposes. They are practice environments where calibration can be built without capital at risk. They cover questions the money venues do not list, which makes them a source of probability estimates on topics that otherwise have none. And they publish their methodology. The question-writing and resolution-criteria standards developed on forecasting platforms (Metaculus's question-writing guide is a widely used reference) are among the clearest articulations anywhere of what makes an event contract well specified.

How do prediction market venue types compare?

Set side by side, the types trade off against each other along a handful of durable dimensions.

Custody and counterparty. Regulated exchanges hold dollars under derivatives customer-protection rules. Crypto-native venues hold tokens in wallets, where protections come from smart-contract design and the venue's own controls rather than a regulatory framework. Brokerage access adds an intermediary account in front of an exchange, and forecasting platforms hold nothing at all. What failure would look like differs accordingly, and so does the available recourse.

Resolution authority. Regulated exchanges settle according to written rules with a regulator behind them, crypto-native venues settle through bonded proposals and dispute votes, and play-money platforms lean on creators and staff. No system removes the hard problem, which is writing questions precisely enough that the answer is checkable. Ambiguous wording produces contested resolutions under every authority structure, and Rethink Priorities' taxonomy of specification problems catalogs the ways questions go wrong in terms that apply to all of them.

Catalog formation. Regulated listing processes tend to produce narrower catalogs vetted for compliance. Permissionless or lightly gated creation produces breadth with variable question quality, and brokerage catalogs are curated subsets of an exchange's list. Whether a question can trade anywhere at all is often determined by this dimension before any other.

Capital mechanics. Binary contracts are fully collateralized everywhere, so capital sits inside positions until resolution. Venues differ in whether that collateral earns interest, a detail that matters most for long-dated contracts. When locked capital earns nothing, prices on far-out questions can drift from the probabilities traders actually believe, because holding the position has a carrying cost (a well-known analysis of long-dated markets works through the distortion).

What the price means. Every type outputs a probability-shaped number, whether as a real-money price, a play-money price, or an aggregated forecast. A long research literature treats these numbers as usable probability forecasts across very different market designs (Wolfers and Zitzewitz's survey is the standard starting point), and cross-platform tools lean on that equivalence in practice. Aggregators such as ElectionBettingOdds average betting-exchange odds and market prices into a single probability per outcome, and calibration trackers such as Calibration City compare how well each platform's numbers line up with realized outcomes. The incentive structures differ, and real-money prices tend to draw the most scrutiny. The shared output format is what makes venue types comparable at all.

Which structure fits depends on what the participant is doing. Regulated exchanges suit those who want dollars in, dollars out, and a rulebook with a regulator behind it. Crypto-native venues suit those who want breadth and self-custody and are comfortable with tokens and oracle-based resolution. Brokerage access suits those who want event contracts next to an existing portfolio with minimal setup, and play-money and forecasting platforms suit anyone building the underlying skill, which transfers across all of them. The prices these venues produce all answer to the same interpretive toolkit, from reading them as probabilities to understanding how crowds aggregate information in the first place.

Related guides