Portfolio & risk intelligence
Shared exposure, drift, decay, incidents, disagreement, lineage and research priority. Information only — Risky Business owns risk.
Portfolio
Open tickets as one structure. Separate metrics — never one blended number. PAPER and REAL never mix.
Runs every 30 minutes in the cloud worker.
Runs every 30 minutes in the cloud worker.
Model version health
Drift opens a review. Nothing is retrained or recalibrated automatically.
Real-data validation is pending.
Recipe version health
Decay needs evidence; a short bad streak never retires a recipe.
Needs settled real-evidence tickets.
Drift & alerts
Deduplicated; lifecycle open → acknowledged → resolved.
kalshi: schema drift
seen 14× · since 2026-10-12 20:16 UTC
Restore drill failed
seen 1× · since 2026-10-12 20:16 UTC
Provider incidents & trust
Incidents persist; trust recovers as incidents age. Trust is a state, not a probability.
recent 0.0% · historical 0.0% · incidents 30d 1 · open 1
kalshi · 2026-10-12 · 14× · 0 records
Cross-provider disagreement
A research signal — disagreement never means the model is right.
0 recent cases · 0 autopsied · SMALL SAMPLE
Research queue
Ranked by Expected Information Value — a bounded heuristic, not a certainty. Operational work always comes first.
- 1. PROVIDER INCIDENT · score 3.194 · EIV 0.50kalshi: SCHEMA_DRIFT
- 2. COVERAGE GAP · score 0.095 · EIV 0.00Close data gap: NBA HISTORICAL_ODDS (UNVERIFIED)
- 3. COVERAGE GAP · score 0.095 · EIV 0.00Close data gap: NCAAF HISTORICAL_ODDS (UNVERIFIED)
- 4. COVERAGE GAP · score 0.095 · EIV 0.00Close data gap: NCAAF RESULTS (UNAVAILABLE)
- 5. COVERAGE GAP · score 0.095 · EIV 0.00Close data gap: NFL HISTORICAL_ODDS (UNVERIFIED)
- 6. COVERAGE GAP · score 0.095 · EIV 0.00Close data gap: NFL RESULTS (UNAVAILABLE)
Deferred: Real-data NBA moneyline validation (DEFERRED DATA); Season-phase regime breakdown (hypothesis) (DEFERRED DATA); First prop model (NBA points) validation (DEFERRED DATA)
Data lineage & contamination
Payload → record → feature → prediction → leg → ticket → experiment → finding. Contaminated results are marked, never deleted.
No contamination marks.
Trace the latest ticket ›Hedge research & stress tests
Research only. A hedge is not assumed to improve EV; protection is never double-counted.