Smart contracts execute deterministic logic: balances, swaps, redemption. Elections, CPI prints, and game scores live off-chain. Oracles are the bridge—mechanisms that import a claimed real-world outcome so tokens can settle. The prior chapter showed creation wiring the oracle adapter; this one explains the bridge itself and why settlement and dispute chapters are really oracle games with extra steps.
You never hold “the event.” You hold tokens whose payoff depends on oracle acceptance. That is why oracle chapters belong after creation: the adapter wired at deploy determines how messy reality becomes redeemable state.
The oracle problem
The world delivers messy facts, revisions, and ambiguity. The chain wants binary state updates. Wire services call races; agencies publish tables; courts delay. Oracles translate headlines into YES, NO, INVALID, or delayed flags traders can redeem against.
Clear resolution rules were always the foundation of trustworthy markets. Honest event probabilities are a separate skill. Oracles demand a third forecast: the probability the on-chain reported outcome matches your reading of reality and the contract text.
Lifecycle from trade to redeem
While trading is open, you size like any market: executable price, fees, slippage. When the real world resolves, marks may gap before the chain agrees. Someone asserts or reports an outcome; a challenge window may open; capital can freeze. After finality, winners redeem collateral; losers go to zero or refund paths on invalid.
Regulated exchanges abstract this to operations—fast, with narrow appeals. Polymarket-style stacks use optimistic oracles: public assertion plus bonds. Augur-style reporting uses stake-weighted rounds—slower, with explicit invalid outcomes common on DIY wordings.
What oracles deliver
Point outcomes pick YES or NO. Scalars return numbers for range markets. Categorical winners select one of several mutually exclusive legs. INVALID or void means no clean mapping—refund rules apply instead of a heroic win. Delayed flags extend decay when data never arrives on time.
Each product shape—binaries, scalars, trees, bundles, conditionals—changes what the oracle must output, not just how loud the headline is.
Clean feed versus wording fight
March CPI year-over-year prints 3.1% against a 3.0% threshold. Spot YES jumps toward ninety-five cents. An asserter posts YES to an optimistic oracle; no one challenges within the window; finalize redeems YES at one dollar. Oracle risk was small because rules cited the official series.
Now take: “Will Candidate X win the debate per post-debate poll average?” Averages split forty-nine to fifty-one. Holders argue “win” means plurality of snap polls. Bonds escalate; mids whipsaw; your seventy-cent YES can path to zero or one over weeks. Props are a different asset class than agency macros—price dispute gamma, not just event risk.
Invalid as a first-class outcome
“Will Product Y launch by Q2?” ships a beta labeled preview. Commercially you were right that it is not a full launch; contractually reporters may call INVALID and refund near entry rather than pay NO. You were right about the world and flat on the wallet. Model invalid probability before entry, not as a post-hoc excuse.
Two probabilities traders confuse
Your world forecast is one number. Implied chance the contract pays your token is another—shaped by assertion risk, invalid paths, and freeze calendar. Ignoring assertion politics overpays when rules favor bond lawyers. Ignoring invalid on permissionless DIY props overpays when refund is likely. Ignoring freeze treats capital as free while challenge windows run.
Cross-venue consensus on price only helps when rules and oracle families match—not when one mid is thin on a decentralized venue and another is deep on a hybrid with different appendices.
Sources oracles trust
Official agencies and series IDs anchor macro markets. Wire and media calls trade speed against certificate-style rules. On-chain-native inputs work when the definition is truly on-chain; wrapping elections still needs a human layer eventually. Human panels add flexibility and politics. Social headlines are fast and ambiguous.
Reporting is a scoring game for asserters and voters maximizing stake, not your profit and loss.
Practical discipline
Before size: classify resolver type, estimate dispute and invalid priors from wording and category history, calendar freeze days as opportunity cost, and fold tails into expected value and Kelly fractions. Pre-assert exits often dominate holding conviction through bond fights. Headlines are not oracle outputs; rules are.
After final, journal separately: were you right about the world, and what did the contract pay?
Timing: headlines versus finalize
Election night is emotionally loud; oracle finalize may be quiet and days later. Marks can pin at ninety-eight cents while challenge windows remain open—that discount is often dispute risk, not free money. Macro prints can gap instantly when asserters race to post the obvious outcome; props linger in limbo. Build a calendar column for assert and challenge deadlines beside your event calendar.
When restatements do not matter
Agencies revise data months later. Most event contracts do not reopen because a series was restated unless the rules explicitly say so. Your oracle moment was the publish window defined at creation. That immutability helps finality and hurts traders who confused “truth eventually” with “contract will update.”
Scalar and categorical oracle demands
A range market needs a number, not just YES or NO. A four-way election needs exactly one winner token honored. If the oracle posts the wrong slice or delays while “Other” is undefined, settlement fails in ways mids never warned. Product shape at creation tells you what the oracle must output—read it before you trade the headline.
Common mistakes
Conflating cable news with settlement. Zero dispute prior on props. Holding through bond fights for “justice.” Treating invalid as a glitch. Using one world probability to arbitrage across oracle families.
Synthesis
The oracle is the settlement layer. Trading skill selects entry; oracle skill selects payout. Build habits: read rules first, model dispute and invalid, calendar challenges, exit when the fight is priced, journal world versus chain outcomes separately.
What comes next
Oracles translate messy reality into contract state; the next chapter classifies who is allowed to speak that truth.
Feeds are not oracles until rules say so
A wire call is not settlement until the contract’s oracle path accepts it. Traders who live on Twitter learn this on assertion day. Build a habit: when news breaks, ask what has to happen on-chain before redeem—not what has to happen on television.
Practice note
Add assert and challenge dates to your trading calendar for any open crypto position. Treat those days like earnings for a stock—liquidity and mark meaning can change completely.
Reader takeaway
Headlines resolve in media; contracts resolve on oracle timelines. Model dispute, invalid, and freeze as part of expected value. Journal whether you were right about the world and what the chain paid. The next chapter sorts oracle families so you can choose venues with open eyes.
Next: Oracle Types: Centralized, Decentralized, and Cryptographic