Prediction Market Oracle System
Prediction market relies on trust in oracle: if oracle lies or mistakes — entire market becomes pointless. Oracle system for prediction market more complex than price feed — must handle subjective questions, disputed outcomes, force majeure, events that didn't happen.
Market Classification by Oracle Requirements
Binary markets (yes/no): simplest resolution. "BTC price > $100K on Dec 31?" → 0 or 1.
Scalar markets: "What will ETH price be Dec 31?" → numerical value in range.
Categorical markets: "Who wins championship?" → choice from several options.
Conditional markets: "If event X happens — what with Y?" — complex dependency.
Multi-Tier Oracle Stack
Polymarket-style platform uses UMA Optimistic Oracle for most markets plus USDC on Polygon.
Three-Tier System
Tier 1: Automatic Resolution
├── Price feeds (Chainlink/Pyth) for price-based
└── Verifiable external data (sports, election APIs)
Tier 2: Optimistic Resolution (UMA/Reality.eth)
├── Proposer → propose outcome + bond
├── Dispute window (2h - 24h)
└── No dispute → accepted
Tier 3: Human Escalation
├── UMA DVM (token voter court)
├── Kleros arbitration
└── DAO governance vote
Market Creator Role
Each market created with:
- Resolution criteria: exact outcome-determining conditions
- Resolution oracle: which oracle/process used
- Resolution timestamp: when or how triggered
- Invalid conditions: when market recognized invalid
Trusted Reporters / Verified Sources
For sports and news events — trusted reporter model:
News Feed Integration: official API (AP, Reuters, ESPN) provide machine-readable data. Contractual agreements ensure source reliability.
Multi-reporter consensus: several independent reporters must agree. 3-of-5 threshold.
Reporter staking: reporters stake tokens. Wrong or manipulated report → slashing. Economic incentive for honesty.
Edge Case Handling
Event cancelled: match rescheduled, elections cancelled. Market should refund stakes (void).
Ambiguous outcome: results can be interpreted differently. Need escalation mechanism.
Late oracle data: oracle returned data 2 hours after deadline — accept or not? Criteria set beforehand.
Oracle manipulation: high UMA bond means manipulation uneconomical. Need monitoring system.
Resolution Analytics
Platform should track resolution quality:
- Time from resolution timestamp to actual resolve
- % of markets auto-resolved vs disputed
- Dispute history and outcomes
- Reporter reputation scores
Well-tuned oracle system — competitive advantage for prediction market. Users trust platform with transparent resolution history. Development: 6-10 weeks.