Range markets: trade where a price lands

From threshold events to ranges

A range market with N ranges is built from N−1 ordinary yes/no Novak events — the boundaries — with ascending thresholds: boundary i is “price ≥ threshold_i at time T?”. The winning range is simply the number of boundaries that resolved TRUE. Each boundary goes through the normal resolver quorum and, if challenged, the committees — and any other market can reuse the same boundary events.

Pricing: the LMSR market maker

Let q_i be the shares outstanding in range i and b the liquidity parameter. The market maker charges according to the cost function

Choosing b

Larger b = deeper market (smaller price moves per trade) but a bigger subsidy for the creator.

Lifecycle

1. 1

   Create

   creator
   Create the boundary ladder in one transaction (createEvents), approve USDG, then createMarket(question, boundaryIds, tradingClosesAt, b). The contract pulls ⌈b·ln N⌉ + 1 units and every range starts at 1/N.
2. 2

   Trade

   anyone
   Buy and sell any range. Every trade emits Trade(…, pricesAfter), which the app uses for the probability history.
3. 3

   Trading stops

   automatic
   At tradingClosesAt — and also the moment any boundary is decided, so nobody can trade on a known outcome (the same fix as the yes/no market's fund-loss bug).
4. 4

   Settle

   keeper
   settle(market) reads every boundary in one call through Settlement.getSettlements. Fees go to the TreasuryVault tagged with the boundary events; leftover subsidy goes back to the creator.
5. 5

   Redeem

   traders
   redeem(market) pays each wallet once: 1 USDG per winning share, or 1/N per share if voided.

Settlement rules

Why it can always pay

The market keeps a reserve: the creator's subsidy plus every net trade (fees are held separately). The subsidy is exactly C(0) = b·ln N, so after any trades the reserve equals C(q). For the LMSR, C(q) ≥ max q_i ≥ average q_i — enough to pay the winning range in full, or 1/N on every share if voided. The contract doesn't rely on the proof alone: every buy and sell re-checks reserve ≥ C(q) and reverts otherwise, and rounding always favours the market. Foundry fuzz tests run random trade sequences against every outcome.

Insights: buy, hold or avoid

Each market page compares the market's price for every range with a simple model: a lognormal distribution centred on the live Chainlink spot price, using that feed's own realized volatility and the time left until T.

Create one

In the app: Markets → Create market → Price range — pick any of the 34 Chainlink stock/SGOV feeds, a time T and a liquidity; the ranges are centred on the live price. Or with the SDK:

import { NovakClient, buildPriceLadderSpecs, ladderBucketLabels, feedBySymbol, getDeployment } from "@novakoracle/sdk";

const dep = getDeployment(46630);                       // Robinhood Chain testnet addresses
const client = new NovakClient(publicClient, walletClient, dep);
const nvda = feedBySymbol("NVDA")!;                     // Chainlink feed on Robinhood Chain mainnet
const at = BigInt(Math.floor(Date.now() / 1000) + 48 * 3600);
const thresholds = [230n, 232n, 234n, 236n, 238n].map((d) => d * 10n ** 8n); // 8 decimals

// 1. Five boundary events "NVDA >= $X at T", one transaction
const specs = buildPriceLadderSpecs({ feed: nvda.address, thresholds, at, maxStaleness: 4n * 3600n });
const tx = await client.createEvents(specs, account);
const boundaryIds = await client.getCreatedEventIds(tx);

// 2. Fund the LMSR subsidy and open the market
const b = 500_000_000n;                                 // 500 USDG (6 decimals)
const approve = await client.approveCollateralFor(dep.distributionMarket, client.distributionSubsidy(b, 6), account);
await publicClient.waitForTransactionReceipt({ hash: approve });
const mtx = await client.createDistributionMarket("Where will NVDA be at the bell?", boundaryIds, at - 60n, b, account);
const marketId = await client.getCreatedDistributionMarketId(mtx);

console.log(ladderBucketLabels(thresholds));            // ["< $230", "$230–$232", …, "≥ $238"]

Contract reference

Like every consumer, DistributionMarket reads outcomes only through Settlement → IEventBus — enforced by test_holdsOnlySettlementReference.