Market shapes: binary, labelled, multi-outcome, price bucket, scalar

The shape of a market tells you what its price means. It also decides which conversions apply and how the market settles.

A Yes at 0.30 on a binary reads as about a 30 per cent chance. On a scalar band from 50 to 150, a Long at 0.30 means the market expects a settle near 80: 50 + 0.30 * (150 - 50) = 80.

Every shape reuses HIP-4’s one primitive, an outcome paying f and 1 - f, with no new contract type.

Two containers make five shapes

• Standalone outcome. One Yes and No pair on one book.
• Question. Outcomes, each on its own book with sides Yes and No, of which exactly one settles to 1. The protocol adds a fallback to the named outcomes. Outcomes added while it is live credit fallback Yes holders an equal Yes, so “none of the listed” keeps its meaning.

Shape	Container	What it is
Price binary	standalone	the protocol’s recurring markets, named Recurring; Yes if the mark price, interpolated to expiry, is at or above the target. Deployers list the same idea with the binaryPrice template, resolved on a TWAP over a stated number of seconds
Labelled binary	standalone	a template names its sides, such as template:Over and template:Under; the first is side 0 and receives f. Most standalone deployer templates on mainnet, such as sportsOverUnderMarket and aiModelHeadToHead, produce it
Multi-outcome question	question	named outcomes are the answers, such as a tournament winner. On deployer questions the fallback is named template fallback, described other, and settles to 0 automatically when a named answer wins
Price buckets	question	N thresholds cut the price into N + 1 half-open buckets, lower bound inclusive (< P1, [P1, P2), >= P2), plus the fallback
Scalar	standalone	settles to a fraction; on scalarPrice the sides are Long and Short

Example descriptions:

class:priceBinary|underlying:BTC|expiry:20260503-0600|targetPrice:78213|period:1d
class:priceBucket|underlying:BTC|expiry:20260507-0745|priceThresholds:81538,81783|period:15m

The scalarPrice payout:

Long  = clamp((price - low) / (high - low), 0, 1)
Short = 1 - Long

price = TWAP over {seconds} of a Hyperliquid perp's price, ending at {time}

With low 50 and high 150, a price of 116 pays Long 0.66 and Short 0.34. The perp may be a HIP-3 dex:COIN, so a scalar band can list against a perp its deployer does not own (how the HIPs compose).

Every outcome trades on one merged book. Buying Yes at p is selling No at 1 - p. Priority is price, then side, then time, so at one merged level resting sells match before resting buys; measured on mainnet, the Yes and No l2Books were exact 1 - p mirrors.

The numbers

Parameter	Value
Price grid	steps of 0.00001, from 0.00001 to 0.99999
Size decimals	0 (whole contracts)
Named outcomes per question	at most 100, plus 1 automatic fallback
Standalone settlement	any fraction from 0 to 1
Question outcome settlement	exactly 0 or 1
Display name of a standalone outcome or question	at most 70 characters
Display name of a question outcome	at most 25 characters
Display side name, standalone	at most 10 characters
On-chain name / description	at most 100 / 2,000 characters
Price buckets	N thresholds give N + 1 buckets, plus the fallback
Scalar Long payout	clamp((price - low) / (high - low), 0, 1)
Scalar Short payout	1 - Long

Reading prices and placing orders

Shape	A Yes price of 0.30 means
Price binary, labelled binary	about a 30 per cent chance the first side wins
Multi-outcome, price bucket	about a 30 per cent chance this answer or bucket wins; the Yes prices across the question should sum to about 1
Scalar	the market expects Long to pay 0.30, which maps to a price of low + 0.30 * (high - low)

When the best Yes asks across a question sum below 1, buying one of each and redeeming the set pays the difference (split, merge, negate).

Orders follow spot within the grid: a market order is FrontendMarket at 0.99999 to buy or 0.00001 to sell. FOK and GTD are rejected; Ioc fills what it can and cancels the rest (order types).

Classifying an outcome from code

Each outcome in outcomeMeta carries name, description and sideSpecs. Deployer markets are named template:<template id>, with keyword values sorted by keyword as keyword:value|keyword:value. Standalone sides are template: plus the template’s side name; question outcomes use Yes and No.

Classify an outcome by checking in this order, as one production deployer does:

Check	Shape
Description starts class:priceBinary	price binary (protocol recurring)
Parent question’s description starts class:priceBucket	price bucket
Belongs to any other question	multi-outcome
Anything else	labelled binary

Scalars, being standalone pairs, fall in the last row: tell them apart by template name, such as template:scalarPrice or template:sportsScalarMarket.

• Side index is position, not name. Side 0 is the first sideSpecs entry; the encoding is 10 * outcome + side (asset ids). One SDK maps Yes to 0 and any other name to 1, wrong for labelled sides; code that falls back to side 0 on unparsable input trades the wrong side silently, so validate the side.
• Question outcomes carry only their own keywords, such as choice:A; event and deadline times live on the parent question.
• Count the pair as one market. dayBaseVlm is identical on the Yes and No coins, and their two dayNtlVlm values sum to it (measuring volume).
• Open interest is circulatingSupply per side; the sides differed on 89 of 192 pairs measured.

Unresolved rules and network differences

• The spot five-significant-figure price rule does not apply. One SDK applies it, allowing 0.000001 steps below 0.10 that the 0.00001 grid lacks (precision).
• The minimum order size is unresolved: $10 historically, $1 anticipated, neither stated in the official docs for outcomes. One SDK values an order on its cheaper side, min(p, 1 - p).
• A settled outcome leaves outcomeMeta with its bucket bounds; only the parent question’s description in settledOutcome recovers them.
• Template sets differ by network: on 2026-09-14 mainnet had 18 and testnet 60. Mainnet has scalarPrice; testnet lacks it and carries other scalar ids, including a misspelt scalePrice; test against the mainnet id (templates).
• Some templates settle a draw or a void at 0.5, even in a binary shape: read the template before reading a price as a probability.

For where shapes come from, see Outcome templates and keywords.