Frequently asked questions

Chainlink is Robinhood Chain's price oracle, and Novak resolvers read its feeds — it's a data source, not a competitor. Price feeds answer "what is the price now"; Chainlink's own docs say they don't provide corporate-action calendars or pause triggers. Novak turns questions like "was NVDA ≥ $230 at the close?" or "did TSLA's multiplier change this week?" into finalized, disputable facts that every protocol reads the same way.

An optimistic oracle accepts a single bonded proposer's answer and, for hard disputes, escalates to a vote weighted by governance-token holdings — which is how a $200M+ Polymarket market (the Zelenskyy suit market, 2025) was decided against widely reported facts. Novak starts from a quorum of resolvers with evidence, escalates to randomly drawn resolver committees, and if they can't agree the event is Voided and markets refund. It's deliberately narrower: machine-checkable facts about stock tokens, not arbitrary questions.

Resolution and disputes are the expensive, trust-critical part of running a market. A market can create (or reuse) Novak events and settle against EventBus.getAvailability / readOutcome — on-chain with one interface, or off-chain with the SDK's waitForOutcome — and get quorum, disputes, committees and the VOID path for free. See Integrate your market.

A market on where a price lands, split into up to 10 ranges with their own shares. An LMSR market maker keeps the range prices summing to 1, so the chart is the market's probability distribution. It settles on N−1 ordinary threshold events: the winning range is the number of boundaries that resolved TRUE.

You're never paid on a guess. In a yes/no market every depositor gets their own stake back. In a range market every share redeems at 1/N USDG, whatever range it's in. The same happens if a range market's boundaries come back inconsistent.

No. Trading stops at the market's close time and also the moment any of its events is decided — this closed a real fund-loss bug where losers could withdraw after the outcome was public.

No. They compare the market's price with a simple lognormal model around the live Chainlink price, using the feed's realized volatility and the time left. They flag gaps over 5 points. The model knows nothing about news or earnings.

Yes. Markets are enumerated on-chain on Robinhood Chain testnet and prices come from the live Chainlink feeds on mainnet. Collateral is MockUSDG — test money with no value; mint it free in the app.

Three authorized resolvers run today (quorum 2 of 3), all operated by the team. The quorum and committee rules are designed for many independent operators; adding them is the next step. With 7 or fewer resolvers a committee is the whole pool; above that, committees are drawn by commit-reveal.

Resolvers observe shortly after the event time; the dispute window is set per event (a few minutes in the demo markets; the app default is 10 minutes); then the keeper finalizes and settles automatically. A disputed event adds up to 1 hour (Tier 1) and 2 more hours (Tier 2).

Every market fee goes to the TreasuryVault, attributed to the events the market settled on, and is split in thirds per event: resolvers who reported the final outcome, committee members who voted it (or the insurance reserve if undisputed), and the treasury. Wrong or silent resolvers earn nothing from that event.

No. The bonded two-tier committee ladder replaced single-owner arbitration and never falls back to a token vote, committees are drawn by commit-reveal among resolvers rather than an external randomness beacon (Chainlink VRF isn't available on Robinhood Chain), bonds are flat, and there is no staking or reputation yet. See the threat model.

Robinhood Chain testnet (chain ID 46630). Resolvers read their data — Chainlink stock feeds, ERC-8056 stock tokens, Robinhood's asset registry — from Robinhood Chain mainnet, read-only. Local anvil (31337) is supported for development.

No. Everything is read on-chain (or through the SDK). The Contract API page documents the surface with real function signatures; its REST examples are illustrative of a future indexer.

Out of scope for this pass and tracked as a post-MVP extension. AND / OR / NOT / BEFORE / WITHIN cover the current markets.

No. It has 138 Foundry tests (unit, integration, fuzz and adversarial) plus SDK and resolver tests, and a written threat model — but no external audit. It is testnet software.