How PredictHanta resolves hantavirus markets
Every market on the 2026 hantavirus outbreak is settled against a named, verifiable public-health source. This page documents the source hierarchy, the resolution process, the scoring rules used to rank forecasters, and the conflict-of-interest policy that governs our editors.
Operating principles
PredictHanta is an information market on the 2026 hantavirus situation. Markets are not financial advice and the platform does not custody fiat funds. Three principles govern every market we publish:
- Verifiable resolution. Every market specifies, before opening, the exact source URL, agency and document type that will determine YES/NO.
- Calibrated scoring. Forecaster reputation is updated using strictly proper scoring rules (Brier and log score), not P/L.
- Editorial independence. Resolution is performed by editors who are not allowed to take positions in markets they resolve.
Source hierarchy
Markets resolve against a strict hierarchy. Higher tiers override lower tiers if they conflict.
- Tier 1 — Primary public-health bodies: WHO Disease Outbreak News, WHO IHR Emergency Committee statements, ECDC Communicable Disease Threats Report, PAHO weekly bulletins, CDC MMWR.
- Tier 2 — National reference labs and ministries: ANLIS Malbrán (Argentina), ISP (Chile), FIOCRUZ (Brazil), FHI (Norway), RKI (Germany), CDC laboratory confirmation.
- Tier 3 — Peer-reviewed publications: DOI-anchored articles in NEJM, Lancet ID, Nature Medicine, Emerging Infectious Diseases.
- Tier 4 — Surveillance / open data: ProMED-mail, HealthMap, GISAID, GenBank deposits, GDELT event-coding.
We publish the full live source list at /sources.
Resolution process
When a market closes, two editors independently review the designated source. If they agree, resolution settles immediately. If they disagree, the market enters a 24-hour dispute window during which any holder can submit additional Tier-1 / Tier-2 evidence. A senior editor adjudicates within 48 hours of close.
Markets that cannot be resolved on the documented source — for example because WHO did not publish the expected DON in time — are either extended (with explicit notice) or refunded. Refunds are atomic; partial settlements are never issued.
Forecaster scoring
Reputation uses two strictly proper scoring rules computed at resolution time:
- Brier score:
(probability − outcome)², averaged over all forecasts a user has submitted in a market. - Log score:
−log(probability assigned to actual outcome), bounded to[0, 5]to prevent infinite penalties on zero-probability assignments.
Both scores reward calibrated honesty. Lying about your beliefs maximises expected loss under either rule, regardless of trading P/L.
Conflicts of interest
Editors and contributors must declare any consulting, equity, or grant relationship with public-health agencies, vaccine developers, travel insurers or media outlets whose actions could affect a market they touch. Declarations are public on author pages.
PredictHanta itself takes no positions and earns no fee on the information market in this beta. We accept no advertising from parties with material exposure to outbreak outcomes.
Corrections and changes
When we change a market rule after publication, we keep a timestamped diff in the market's resolution log and email all holders. We never silently rewrite resolved markets. Significant editorial corrections to news posts are logged at the bottom of the article.