How Research forward calls differ from the rules-based ledger
AirIndex maintains two distinct prediction surfaces. The rules-based predictions ledger at /predictions surfaces forward signals minted automatically by the auto-classifier pipeline when defined market-signal patterns fire. That’s a high-volume rules-based watch list, structurally patterned on Moody’s CreditWatch.
This page is different. Each entry here is an analyst-grade forward call embedded in a specific published Research piece — a methodology-anchored projection with explicit, falsifiable resolution criteria, issued at human-curated cadence. Lower volume, higher institutional weight. Structurally patterned on a rating agency’s analyst publication track record.
Each call resolves when an event matching its documented criteria is observed in primary-source data within the window — or is marked invalidated, inconclusive, or overdue if the window passes without a qualifying event. Resolution is recorded in PredictionRecord with explicit links to the resolving primary-source record.
The full audit chain — score history, factor breakdowns, classifier outputs, resolution events — is cryptographically anchored. Verifiable via the provenance methodology.
The track record is intentionally narrow today. The institutional pattern is accumulation over time — same way a rating agency’s published accuracy stats compound across hundreds of resolved calls over years. AirIndex Research is in the first months of that accumulation.