What is HIP-4? Prediction markets on Hyperliquid

Contents

What is HIP-4 on Hyperliquid?#

HIP-4 is Hyperliquid's standard for outcome markets: dated contracts that settle inside a fixed range. The HIP-4 page in Hyperliquid's docs describes it as a general building block for prediction markets and bounded options-like instruments, fully collateralized, with no leverage and no liquidations.

HIP-4 markets live on HyperCore, the part of the Hyperliquid chain that runs the order books and the clearinghouse for perps and spot. That makes Hyperliquid's prediction markets different from Polymarket or Kalshi in one structural way: the order book, the matching and the settlement all run onchain, in the same account you use for Hyperliquid perps. For how HyperCore fits into the rest of the chain, see what Hyperliquid is and how it works.

HIP-4 follows HIP-3, which let outside teams deploy their own perpetual markets. The two share a model: Hyperliquid runs the engine, a staked deployer defines the market, and validators can slash the deployer's stake. HIP-3 builder-deployed perpetuals covers the perp side.

How do HIP-4 outcome markets work?#

A HIP-4 outcome has two sides, each a token. The outcomeMeta info request names the sides in sideSpecs, usually Yes and No. You buy Yes if you think the event happens and No if you think it doesn't. Prices run from 0 to 1 USDC, so a Yes price of 0.62 reads as a 62% chance.

Settlement with settleFraction#

When a HIP-4 market settles, the protocol converts every Yes token into settleFraction USDC and every No token into 1 - settleFraction USDC. A binary Yes result is settleFraction = 1, a binary No result is settleFraction = 0. Settlement is automatic: you don't redeem anything.

A standalone outcome can settle to any fraction in [0, 1]. Hyperliquid's HIP-4 deployer actions give "0.66" as an example for scalar payouts. That is how a HIP-4 scalar market pays by how much something happened instead of whether it crossed one line. Scalar settlement on HIP-4 works through the payoff, which is a normalized call spread.

One merged order book for Yes and No#

HIP-4 merges the Yes and No books of an outcome into one. Buying Yes at price p is the same order as selling No at 1 - p, so both sides draw on the same liquidity. Price-time priority becomes price-side-time priority: at the same merged price, resting sells sort ahead of resting dual buys. Anyone measuring depth should count the mirrored Yes and No liquidity once, as Totalis does, because they are the same orders.

You can also convert directly. The exchange endpoint has four userOutcome actions:

HyperEVM contracts can send the same four operations through CoreWriter, the system contract at 0x3333333333333333333333333333333333333333, as action 17, "Outcome operation", per Hyperliquid's page on interacting with HyperCore. How HyperEVM contracts reach HyperCore is covered in HyperCore vs HyperEVM.

Questions: multi-outcome markets#

A HIP-4 question groups several outcomes where exactly one settles Yes, such as "Who wins the tournament?" Each outcome keeps its own order book. negateOutcome and mergeQuestion tie them together, so someone holding No on every outcome can redeem USDC before the question settles.

The deployer rules add detail. Every question gets an automatic fallback outcome named template fallback, the "other" answer. A question holds at most 100 named outcomes. When a deployer adds a new named outcome to a live question, holders of the fallback Yes token receive the same balance of the new outcome's Yes token, so an "other" position keeps its meaning. Question outcomes settle only to exactly 0 or 1. Fractional settlement is for standalone outcomes.

Asset IDs for HIP-4 outcomes#

HIP-4 trading uses the spot API with its own identifiers. Hyperliquid's asset ID docs define encoding = 10 * outcome + side, with side 0 and side 1 the only valid values. The same encoding appears three ways:

Totalis builds its HIP-4 market identity on this numeric encoding, with side 0 as Yes and side 1 as No, rather than on display titles, because titles come from templates and can repeat across deployers.

When was HIP-4 released? The timeline#

Every date below comes from Hyperliquid's official announcements channel unless noted.

Two parts of Hyperliquid's docs lag this timeline as of 2 November 2026. The HIP-4 deployer page is still headed "Testnet-only" and lists testnet limits of 10 active outcomes and 50 deployments a day. The HIP-4 overview page still says outcome fees are zero for initial testing. The announcements above are the more recent primary source on both points.

Who can deploy a HIP-4 market?#

Since the late-August 2026 upgrade, any account that meets the stake can deploy HIP-4 markets on mainnet, but only from templates. Validators vote on templates; deployers fill them in. A template fixes the display text, the side names, the typed keywords and a semanticRestriction that defines what the market may mean. A market that contradicts its template's restriction counts as malformed and is slashable.

The rules from the 20 July announcement and the deployer docs:

Hyperliquid's 14 August announcement said the caps would rise to 1,000 concurrent outcomes and 5,000 deployments a day once the technology stabilized. As of 2 November 2026 we have not found a raised figure in its announcements, so treat higher numbers quoted on social media as unconfirmed.

The 20 July announcement also says that several deployers may list the exact same template instance. Two venues can run the same Fed decision market with separate books and separate settlement.

How does a HIP-4 market get resolved?#

Who settles a HIP-4 market depends on who deployed it:

For recurring price markets the contract specifications define the rule exactly. A binary settles Yes if and only if the mark price, linearly interpolated between the updates just before and just after the settlement time, is at or above targetPrice. A sample description reads class:priceBinary|underlying:BTC|expiry:20260503-0600|targetPrice:78213|period:1d. A price bucket question has three outcomes, < P1, [P1, P2) and ≥ P2, and exactly one settles to 1.

Read each market's rule, not just its title. Totalis's own market data records that the recurring binaries settle Yes at or above the strike, while the deployer binaryPrice template resolves Yes only if the price is strictly above its threshold, measured on a TWAP. One cent can decide which side wins.

For deployer markets, settlement is a signed action with no trader appeal step in the protocol. The details field in settleOutcome must be empty, and since 14 August 2026 Hyperliquid prunes settlement details from L1 state, so a permanent record depends on indexers. The protection is the deployer's slashable stake, not a dispute window. Polymarket's UMA process and Kalshi's regulated determination work differently; the Hyperliquid vs Polymarket vs Kalshi comparison sets them side by side.

What are HIP-4 fees?#

HIP-4 fees follow a different shape from perps and spot. Hyperliquid's fee schedule says outcome trading charges fees only when you close or settle, never when you open:

• Buying a position, or minting a pair, pays no Hyperliquid fee.
• Selling, settling and burning Yes and No pairs pay the fee.
• Outcome markets pay no maker rebates. Accounts that would earn a rebate on spot or perps pay zero on maker orders instead.
• Settlement counts settle_fraction × size toward fee-tier volume.

Deployers add a multiplier. Each template instance carries a deployerFeeScale from 0 to 10, and users pay the base outcome rate times scale + max(scale, 1). The deployer keeps the scale part and the protocol the rest:

The fees page prints no separate outcome tier table. Hyperliquid's 14 August announcement said validator-deployed outcome markets would get fees "on the next network upgrade", and that the average outcome fee would be half the fee on non-outcome spot. The base spot rate is 0.070% taker and 0.040% maker. DefiLlama's Hyperliquid Outcomes methodology dates the switch: Hyperliquid charged no outcome fee of its own before 15 August 2026. As of 2 November 2026 DefiLlama counts $103,420 of HIP-4 fees in the last 30 days, builder fees included, against $79.5 million of volume.

Apps add builder fees on top. HIP-4 builder codes work on sell orders as in spot, and unlike spot they also work on buys, charged in the quote token. Single-market orders in the Totalis app that carry the Totalis builder pay a 0.1% builder fee on filled value, buy or sell; some orders go without it, such as a new account's first buy. Every charge is listed on the Totalis fees and limits page. Hyperliquid builder codes explains the mechanism and Hyperliquid fees puts outcome fees next to the perp and spot schedules.

What markets trade on HIP-4?#

HIP-4 markets fall into the families Hyperliquid's validators have approved as templates, plus the protocol's recurring price markets. Hyperliquid's docs show only a few template IDs as examples (binaryPrice, policyRateDecision, policyRateDecrease), so the list below comes from Totalis's own capture.

On 2 October 2026 Totalis captured Hyperliquid's public outcomeTemplates and outcomeMeta data on both networks:

These are counts of outcomes, not questions or liquid markets, and they are a dated snapshot, not today's total. The mainnet templates Totalis reviewed in that capture cover these families:

Template text differs between networks in ways that matter. In that capture, the mainnet binaryPriceExternal template added a fallback that resolves to 0.5 if no settlement price can be verified 24 hours after the market's time. The testnet version had no such cutoff. A market tested on testnet is not proof of how the mainnet version settles.

Volume is still small. As of 2 November 2026, DefiLlama counts $79.5 million of HIP-4 trading over the last 30 days, $31.4 million over 7 days and $312.0 million since launch. It values each fill at price times shares, counts each trade once, and leaves out splits, merges and settlements because they are not trades. Totalis excludes the same three ledger events from its price charts for the same reason.

Can US users trade HIP-4 markets?#

No, not through Hyperliquid's own interface. The Hyperliquid interface terms, last updated 15 June 2026, bar "Restricted Persons", including anyone residing in or located in the United States or Ontario, Canada, or in sanctioned jurisdictions. They add a separate "Excluded Persons" rule that lets the operator block the outcome markets tab in further jurisdictions. Section 1.3 of the same terms states that neither the interface nor Hyperliquid is licensed by any regulator. Check your own jurisdiction's rules before you trade.

How Totalis builds Combos on HIP-4#

Totalis, which publishes this guide, trades Hyperliquid's own HIP-4 markets in two ways in the Totalis app. A single is one outcome, traded directly on the HyperCore order book. A Combo, the prediction-market version of a parlay, joins 2 to 5 outcomes into one position with one payout. Market makers quote each Combo through a request for quote, and the position sits in the Totalis vault on HyperEVM until Hyperliquid's results settle it. Not every HIP-4 market takes Combo positions; the API rejects a leg whose market takes none.

Building on HIP-4 taught us a few things that the docs don't spell out:

• A Combo multiplies leg results. Each leg pays its HIP-4 result as a fraction, and a Combo pays its payout times the product of those fractions. Two legs that each settle at 0.5 pay a quarter of the payout before fees. It is not a refund. How a Combo settles shows the formula and every settlement state.
• Questions limit which legs combine. A HIP-4 question settles exactly one outcome to Yes, so Yes on two outcomes of one question can never both win. Totalis tells apps to block that pair.
• Result, settlement and payment are separate events. A HIP-4 market resolving on HyperCore, the Combo settling, the win landing in your vault balance and a withdrawal are four steps in Totalis's accounting, even when they happen close together. The vault reads native outcome results on HyperEVM through a HyperCore read precompile.

Combo fees are separate from HIP-4 fees: 1.5% of stake on entry and 1% of any proceeds above the net stake when a position settles or is sold back, per the Totalis fees and limits page linked above. For why multi-leg positions matter in prediction markets, see how prediction market parlays work, and for how makers back many Combos with less locked capital, see collateral efficiency for Combos.

Risks and open questions#

HIP-4 is six months old on mainnet, and the trade-offs are real:

• Thin liquidity. As of 2 November 2026, HIP-4's $79.5 million of 30-day volume is under 1% of the $16.4 billion DefiLlama counts for Kalshi over the same days. Wide books cost more than any fee schedule.
• Settlement trust sits with the deployer. Deployer markets settle when the deployer signs. The protocol gives traders no dispute window; validators can slash the stake after the fact.
• Duplicate markets split liquidity. Nothing stops two deployers listing the same event, each with its own book and its own settlement.
• Template wording decides payouts. "Above" and "at or above", or a 0.5 fallback, change who gets paid. Read the template's description, not the title.

Frequently asked questions#

Can you lose more than you put in on HIP-4?#

No. HIP-4 outcomes are fully collateralized, with no leverage and no liquidations. The most you can lose on a Yes or No position is what you paid for it, plus fees.