Appendix B: Oracle Weights
MAGMA uses category-specific weighted consensus across multiple independent oracle sources. The exact weight assigned to each source per category is not published — publishing these weights would let an adversary identify which sources to target for manipulation.
What Is Documented
| Item | Value |
|---|---|
| Resolution confidence threshold | 70% minimum across active sources |
| FAST narratives (MARKET, 5-minute) | Price-feed oracles only; the full AI reasoning stack is bypassed for speed |
| Consensus Oracle categories (SOCIAL, CULTURAL, CONVICTION) | Incorporate community signal as an additional input |
| Universal fallback | Switchboard, when primary sources are unavailable or return low-confidence data |
Each narrative category routes to the sources most appropriate for its resolution criteria. See Oracle Architecture for the full list of sources used per category.
Approximate Public Weights
The values below mirror the public table in Oracle Architecture. They show which sources are active per category and their approximate relative contribution. Exact configured weights are withheld for the security reasons described under Why Weights Are Not Public.
| Category | Pyth | LLM reasoning | AI web search | AI validation | Switchboard | Specialist Source |
|---|---|---|---|---|---|---|
| MARKET | 40% | 25% | 15% | — | 20% fallback | — |
| SPORTS | — | 20% | 30% | — | — | The Odds API 50% |
| ESPORTS | — | 40% | — | — | — | PandaScore 60% |
| LEGAL | — | 40% | 60% | — | — | — |
| POLITICAL | — | 35% | 65% | — | — | — |
| SCIENTIFIC | — | 40% | 60% | — | — | — |
| GEOPOLITICAL | — | 35% | 65% | — | — | — |
| ECONOMICS | — | 35% | 40% | 25% | — | FRED/BLS APIs |
| TECHNOLOGY | — | 35% | 40% | 25% | — | — |
| SOCIAL | — | 20% | — | — | — | Neynar 50%, Twitter 30% |
| CULTURAL | — | 30% | — | 35% | — | Community 35% |
| CONVICTION | — | 30% | — | 35% | — | Community 35% |
| COMMODITIES | — | 25% | — | — | — | Alpha Vantage 75% |
| MACRO | — | 30% | — | 25% | — | FRED/BLS 45% |
| EQUITIES | — | 30% | 20% | — | — | Alpha Vantage 50% |
| FOREX | — | 15% | — | — | — | Alpha Vantage 85% |
| ECOSYSTEM | — | 30% | 30% | — | — | DeFiLlama 40% |
| ONCHAIN | — | 30% | — | — | — | Dune Analytics 70% |
| BUILDER | — | 35% | 40% | 25% | — | — |
| CAST | — | 30% | — | — | — | Neynar 70% |
| HEALTH | — | 40% | 60% | — | — | — |
| CLIMATE | — | 35% | 65% | — | — | — |
| BUSINESS | — | 35% | 40% | 25% | — | — |
| SATIRE | — | 50% | 25% | 25% | — | — |
Weight Governance
| Property | Detail |
|---|---|
| Storage | Category-specific weights live in protocol configuration. |
| Adjustment | Can be changed by super admin without a contract deploy. |
| Timelock | Subject to the standard admin parameter timelock (72 hours on mainnet). |
| Auditability | Any weight change is logged on-chain. |
Why Weights Are Not Public
Once an attacker knows that source X carries Y% weight in category Z, they can focus their manipulation effort precisely on source X rather than needing to corrupt all sources simultaneously. The security benefit of keeping the exact weights non-public outweighs the transparency benefit of publishing them.
Related Pages
- Oracle Architecture — sources, M-of-N consensus, confidence threshold, and the 48-hour challenge window.
- Narrative Categories — the category taxonomy these weights apply to.