← All posts

Prediction Markets on Hyperliquid: Complete HIP-4 Guide

How Hyperliquid prediction markets work in 2026: HIP-4 outcome contracts, order books, settlement, market types, risks, and how to trade with Purrdict.

What are prediction markets on Hyperliquid?

Prediction markets on Hyperliquid are bounded-payout outcome contracts that trade through HyperCore’s central limit order book. HIP-4 is the name commonly used for this outcome-trading system. A contract normally settles to 1.0 when its side wins and 0.0 when it loses, so its market price can be read as a probability-like signal.

That does not make the displayed price a guaranteed probability. It is still an executable market price shaped by bids, asks, spread, depth, inventory, and information. A thin contract offered at 0.60 may be much harder to buy at size than a deep contract at the same displayed price.

This guide focuses on using Hyperliquid prediction markets: what the contracts represent, how orders and settlement work, which market forms exist, what can go wrong, and where Purrdict fits. For the identifier math, observed action types, and deeper protocol details, use our canonical HIP-4 protocol guide.

Are HIP-4 markets live on mainnet?

Yes. Curated HIP-4 outcome markets trade on Hyperliquid mainnet. Purrdict’s main trading interface discovers and displays those live markets at app.purrdict.xyz.

There is also a separate permissionless testnet track. On testnet, an eligible account can activate an outcome deployer and create supported market templates through protocol actions. Testnet funds have no monetary value, and permissionless testnet launch should not be described as permissionless mainnet launch.

The distinction matters because old articles and screenshots often collapse three different things:

  • curated mainnet outcome markets with real collateral;
  • permissionless testnet deployment using faucet assets;
  • product experiments, such as Purrdict’s parlay escrow, that are separate from native HIP-4 books.

This page states the network whenever a feature is not mainnet.

How a Hyperliquid prediction market trade works

Each trade has three layers: the market specification, the outcome contract, and the order book.

The market specification tells you the question, expiry, resolution source, and any void or tie rules. The outcome contract identifies the side you are buying. The order book contains resting bids and asks for that side.

Suppose a Yes contract is offered at 0.64. Buying 100 contracts costs 64 units of the market’s collateral before applicable fees. If Yes settles at 1.0, the gross settlement value is 100 and the gain before fees is 36. If Yes settles at 0.0, the contracts are worth zero and the 64-unit purchase cost is lost.

The useful formulas are:

maximum gross value = contracts × 1.0
purchase cost       = contracts × execution price
maximum gain        = contracts × (1.0 − execution price)
maximum loss        = purchase cost

The maximum loss is bounded when you buy an outcome contract outright. That is different from saying the trade is safe. You can still lose the full amount paid, cross a wide spread, trade the wrong side, misunderstand the resolution rules, or hold an illiquid position you cannot exit at a reasonable price.

HyperCore order books instead of an AMM

HIP-4 sides trade through a central limit order book rather than a constant-product prediction-market AMM. Traders post limit orders at chosen prices. Aggressive orders fill against the best available resting prices in price-time priority.

This has several practical consequences:

  • the last traded price is not necessarily the price available now;
  • a market order can fill across several price levels;
  • a tight top-of-book spread can hide weak depth behind it;
  • two related sides do not have to sum to exactly 1.0 at executable size;
  • tick size and minimum size rules must be respected when constructing orders.

Before buying, inspect both the spread and the quantity available at each level. A probability thesis can be right while the trade is bad because the entry price or slippage is wrong.

HIP-4 market types

Observed HIP-4 markets fall into a handful of recurring shapes. Their human-readable descriptions are important because a single type flag does not carry the complete meaning.

Binary outcome markets

A binary market has two mutually exclusive sides, often labelled Yes and No. The question should define exactly what makes either side win. A well-specified price market includes the underlying, threshold, expiry time, and price source.

Multi-outcome questions

A multi-outcome question groups several possible answers. Each answer can be represented through its own tradeable outcome structure. Traders should verify whether the outcomes are mutually exclusive, collectively exhaustive, and whether a fallback or “other” outcome exists.

Do not assume the visible prices across a multi-outcome group add to exactly 1.0. Spread, depth, and independently quoted books can create gaps.

Recurring price binaries

Recurring markets reuse a template for successive time windows. An observed description can encode a class, underlying, UTC expiry, target price, and period. A new round has a new identity even when the displayed question looks almost identical to the previous round.

Always check the expiry. Buying yesterday’s settled or nearly expired round because the label resembles today’s market is an avoidable error.

Price buckets and named outcomes

Price-bucket questions divide a possible settlement price into ranges. Named-outcome markets use human labels rather than generic Yes and No. Both require careful mapping between the displayed label and the actual contract identifier.

How HIP-4 settlement works

When an outcome resolves, winning contracts receive a settlement fill at 1.0 and losing contracts receive a settlement fill at 0.0. Settlement fills are marked with the direction Settlement, and the realized result appears automatically in the account. There is no separate claim transaction in the normal flow.

Current mainnet HIP-4 market metadata and settlement records use USDC. Legacy testnet fixtures may still expose USDH, so integrations should read live asset metadata and balance changes for the selected network instead of carrying an old testnet assumption into mainnet.

For price-based outcomes, the public settledOutcome information endpoint can expose the settled side and recorded oracle-price details after resolution. The vote or action that triggered settlement is not exposed through every public API, which is one reason an archive of L1 actions and settlement fills is useful.

Settlement being automatic does not remove specification risk. Before trading, read:

  • the exact cutoff time and timezone;
  • the data or event source;
  • how equality at a threshold is handled;
  • cancellation, tie, fallback, and void behavior;
  • whether the question has been edited or superseded in the interface.

Understanding HIP-4 identifiers

The same outcome side can appear in several identifier forms. API and WebSocket data commonly expose a #N side coin. Balances can use a +N form. Other metadata and order actions can use pair or integer asset indices.

These forms are related, but they are not interchangeable strings. The integer asset field used by an order is not simply the visible #N label with its prefix removed. A wrong conversion can submit an order for the wrong asset or fail without an obvious user-facing explanation.

Application developers should use tested conversion helpers rather than reimplementing the formulas. Purrdict publishes those helpers in the open-source @purrdict/hip4 TypeScript package.

Fees and trading costs

The total cost of a prediction-market trade includes more than a headline fee:

  • bid/ask spread;
  • slippage across order-book levels;
  • any protocol or builder fee attached to the order;
  • the opportunity cost of collateral until exit or settlement;
  • adverse selection when another trader has fresher information.

Observed HIP-4 builder fees are attached on the sell side and encoded in signed order parameters. Interfaces should show the applicable fee before a user signs. Fee rules and caps can change, so a current order preview is more reliable than a fee number copied from an old article.

How to trade HIP-4 markets with Purrdict

The basic mainnet flow is:

  1. Open Purrdict and connect a compatible wallet.
  2. Choose a market and confirm that the network indicator says mainnet.
  3. Read the full question, expiry, source, and settlement rules.
  4. Select the intended side and inspect the live order-book depth.
  5. Choose a limit price or review the estimated execution range for an aggressive order.
  6. Confirm contract quantity, maximum cost, and the fee shown in the order preview.
  7. Sign the order with the configured Hyperliquid agent wallet.
  8. Monitor the position, sell it before resolution if liquidity is available, or hold it for automatic settlement.

Agent wallets allow the application to sign permitted trading actions without asking for a browser-wallet popup on every order. They should be treated as scoped trading credentials: keep the key local, understand its permissions, and revoke it if the device or session is no longer trusted.

For a screen-by-screen introduction, see how to trade prediction markets.

Launching your own prediction market

HIP-4 testnet includes an observed permissionless Standard-mode deployment path. The current flow requires 100 HYPE staked to a validator, a one-time outcome-deployer activation, a unique permanent venue name, and a supported market template. Current testnet limits include caps on simultaneously active markets and daily launches.

Purrdict’s Launchpad guides the wallet through those actions without asking for a private key. It validates the form and shows the exact action before signature. Read the complete prediction-market launch guide before using it.

This is a testnet workflow. Do not send mainnet funds to a testnet address or interpret a successful testnet deployment as a promise of mainnet listing.

Parlays and combo prediction markets

A parlay combines several outcome legs into one all-or-nothing position. Native HIP-4 order books supply the underlying legs, but the combination is not itself a native HIP-4 market. Purrdict’s executable Parlay v2 design uses a separate RFQ, HyperEVM escrow, and settlement system that is currently testnet only.

This separation is deliberate. Creating a persistent order book for every possible combination would fragment liquidity. A request-for-quote system can price a requested combination just in time, including correlation between its legs.

Learn more in the canonical guide to prediction market parlays and combos and the pooled prediction market liquidity vault.

Risks to check before every trade

Resolution risk

An outcome can be economically obvious while its written rules resolve differently. Trade the specification, not the headline.

Liquidity risk

You may not be able to exit at the displayed probability. Size positions using available depth, not only the last price.

Operational risk

Wrong network selection, an incorrect contract side, stale application data, or a compromised signing key can turn a correct thesis into a loss.

Protocol and application risk

HIP-4 behavior has been documented primarily through observed API and on-chain data rather than a complete official standalone specification. Smart-contract, exchange, data-indexing, and frontend defects remain possible.

Prediction-market rules vary by location and can change. Wallet access does not establish that trading is permitted in a user’s jurisdiction. Users are responsible for determining which products they may use.

Frequently asked questions

Is HIP-4 an official published specification?

Hyperliquid has not published a complete standalone HIP-4 specification covering every observed identifier, action, and settlement behavior. Purrdict’s documentation is based on live API responses, on-chain data, node data, and repeated integration testing. It is an independent reference, not official Hyperliquid documentation.

Can a HIP-4 position be liquidated?

An outright purchased outcome contract has a bounded purchase cost and does not behave like a leveraged perpetual position. The buyer can lose the full purchase cost, but there is no additional loss beyond that position’s paid amount from ordinary 0-or-1 settlement.

Does Yes plus No always equal 1?

At final settlement, one binary side normally receives 1.0 and the other 0.0. Before settlement, executable bids and asks can leave a spread, so the prices available at a chosen size need not add to exactly 1.0.

Are permissionless markets on mainnet?

Curated HIP-4 markets trade on mainnet. The permissionless Standard-mode launch flow described by Purrdict is currently a Hyperliquid testnet feature.

Are Purrdict parlays native HIP-4 contracts?

No. They reference HIP-4 outcomes as legs but use a separate synthetic RFQ and HyperEVM escrow design. The executable system is currently testnet only.

Where to go next

Purrdict is an independent application built on public Hyperliquid infrastructure. It is not operated by or affiliated with Hyperliquid Labs.

Ready to trade?

Explore curated HIP-4 outcome markets on Hyperliquid mainnet through the Purrdict trading interface.

Start Trading → Browse Markets

Ready to trade?

Explore curated HIP-4 markets on Hyperliquid mainnet with Purrdict.

Start Trading →