Code Cannot See Outside the Window

A source, a report, and a final result are different

The prediction market asks: “Will humans return to the Moon before 2028?” The price can move onchain, but whether the event happened must come from outside. A complete oracle flow has at least three layers:

1. Data source: official records, APIs, or evidence that may be delayed, conflicting, or manipulated.
2. Initial report: a participant writes YES or NO and locks a bond.
3. Final result: only after the challenge window and dispute process may a market read the result and pay out.

Why a challenge window matters

If a report settled immediately, a delayed or manipulated source would directly trigger payouts. A challenge window gives other participants a public interval to submit the opposite result and lock a bond that proves they are willing to bear the risk.

This lab simplifies the window to two blocks. Advancing time changes the deadline, not the report. Only after the window ends can the protocol finalize the report or enter arbitration. Real systems may use timestamps, dispute rounds, voting, or decentralized evidence, but the question is the same: when can an error be detected, and why would an honest participant step forward?

A bond makes an opinion accountable

The Primary Reporter and Challenger both lock credits. With no challenge, the reporter gets the bond back, even if the result is wrong; with a challenge, the arbitrator reads the evidence, the correct side receives both bonds, and the wrong side absorbs the loss.

This is not magic correctness. It puts attack cost, discovery reward, and time into the state machine. An attacker may still control a source, collude, or make honest challenges too expensive.

submit(outcome, bond):
  lock(reporter, bond)
  open(challengeWindow)

challenge(oppositeOutcome, bond):
  lock(challenger, bond)

after window:
  if no challenge:
    finalize(reportedOutcome)
  else:
    winner = evidence == reportedOutcome ? reporter : challenger
    winner receives both bonds

Why the market cannot repair a bad fact by itself

An order book knows how many credits traders will exchange for a position; it does not know whether the Moon launch or match actually happened. If an oracle writes the wrong result into settlement, a later price rebound cannot automatically reverse payouts.

That is why serious systems separate price discovery from fact reporting. The first handles beliefs and liquidity; the second handles sources, challenges, and disputes. Clear module boundaries make risks easier to test and explain.

Five questions for evaluating an oracle

After the lab, explain:

1. What is the different role of a data source and a reporter?
2. Why should a report not trigger a payout immediately?
3. Why must a challenger lock a bond?
4. Why can a wrong report still finalize when nobody challenges it?
5. When should a prediction market read the final result?

Oracles are the reality interface for onchain finance: prediction markets, lending collateral ratios, stablecoin reserves, and liquidation prices all depend on them. Understanding their boundary is part of understanding blockchain finance itself.